Big Data and Cognitive Computing doi: 10.3390/bdcc10060173

Authors: Orlando Mauricio Chuquin-Machangara Alex Joel Ajila-Masache Gabriela Abigail Villalta-Jimbo Mario Perez Renato M. Toasa

Open fiscal data in Ecuador remains largely unexplored beyond basic descriptive reporting, despite its potential for territorial intelligence and fiscal planning. This study examines how taxpayers are distributed across Ecuador’s provinces and economic sectors by applying a Big Data pipeline built on Apache Spark 3.5, PostgreSQL 14/PostGIS 3.2, and Python 3.11 spatial libraries to the SRI Tax Registry, comprising approximately 2.5 million records. The analysis combined K-Means and DBSCAN clustering with spatial autocorrelation methods, including Moran’s Index and LISA, to identify concentration patterns and territorial dependencies. The findings show that 68% of taxpayers are located in three provinces, namely Pichincha (34%), Guayas (24%), and Azuay (10%), with a spatial Gini coefficient of 0.61 reflecting considerable fiscal inequality across the country. A Global Moran’s Index of 0.49 (p < 0.001) confirms that neighboring provinces tend to share similar taxpayer densities, while LISA revealed five High–High clusters in major urban centers and six Low–Low clusters in the Amazon region and northern border. DBSCAN identified 27 spatial groupings, including secondary economic nuclei in cities like Ambato, Riobamba, and Machala that autocorrelation models alone do not capture. The methodology is replicable and offers a practical basis for designing place-based fiscal policies in similar contexts. These results provide tax authorities and regional planners with an empirically grounded, scalable framework for identifying territories with fiscal formalization gaps and designing geographically targeted interventions to reduce territorial inequality in Ecuador and in comparable developing-country contexts.

Big Data and Cognitive Computing doi: 10.3390/bdcc10060172

Authors: Andrey Martynov Maria Lapina Mikhail Babenko

This paper addresses the challenge of providing operational access to current metadata in complex, ever-changing relational data warehouses. Traditional catalogs struggle to keep up with changes in schemas, code, and processes. The paper presents a methodological approach based on a dual-loop architecture with ReAct agents and retrieval-augmented generation. The first loop, managed by an Ingestion Agent, continuously updates the semantic layer by automatically analyzing changes. The second loop uses an Assistant Agent to give analysts, developers, and support engineers an intelligent interface. This interface combines semantic search over a vector database with direct execution of diagnostic queries through an extensible set of tools. The main goal is to create a self-updating metadata ecosystem that provides operational access to contextual information for different user groups. The approach’s practical effectiveness is demonstrated through end-to-end scenarios, such as creating complex queries based on business terms or diagnosing extract-transform-load processes.

Big Data and Cognitive Computing doi: 10.3390/bdcc10060171

Authors: Xin He David Camacho Lama Moukheiber Meghna Iyer Benjamin Zhao Christophe Ye Batuhan Nursal Xinyu Guo Albert J. B. Lee Cassie S. Mitchell

Background: Mechanistic reporting from large-scale biomedical knowledge graphs remains challenging, particularly when integrating structured graph evidence with large language model (LLM)–based explanation in a reproducible and auditable manner. Existing approaches either rely on manual synthesis of graph-derived results or generate unconstrained narratives that lack traceability to underlying evidence. Methods: We present SemNet Explorer, an evidence-grounded knowledge graph–LLM unified framework for automated mechanistic reporting across biomedical domains using SemNet 2.0, a PubMed-scale heterogeneous knowledge graph. Given a set of target concepts and a selected semantic layer, the framework organizes graph-derived evidence into structured regions and generates two complementary report types: global reports for process-level mechanisms and anchor-centric reports for localized mediator-based explanations. A central methodological contribution is an ablation-derived adaptive grounding policy: we systematically compare alternative evidence-integration strategies across report types, semantic layers, and region structures, and use the resulting preferences to guide prompt selection in the deployed system. Results: SemNet Explorer produces stable region decompositions and interpretable report scaffolds across molecular (AAPP), disease-level (DSYN), and pharmacologic (PHSU) representations. For global reports, explicit evidence grounding improves expression quality more consistently than content accuracy, with benefits dependent on evidence density and semantic abstraction. In contrast, anchor-centric reports show consistent improvements in both content and expression under stronger, mediator-constrained prompting. These findings are supported by both pairwise ablation comparisons and absolute score analyses. Conclusions: SemNet Explorer establishes a generalizable unified framework and interactive platform for transforming knowledge graph evidence into reproducible mechanistic narratives across biomedical domains, including multimorbidity analysis, comparative pathophysiology, drug repurposing, and adverse event discovery. The results demonstrate that effective knowledge graph–LLM integration requires adaptive, context-dependent evidence grounding rather than fixed prompting strategies.

Big Data and Cognitive Computing doi: 10.3390/bdcc10060170

Authors: Niddal H. Imam

The increasing prevalence of digital communication has made students a primary target for various cyber threats, including identity deception and impersonation techniques that can lead to data breaches and financial loss. In Saudi Arabia, where the youth population is digitally active and integrated into online learning environments, understanding their vulnerability to such threats is paramount. This paper investigates university students’ awareness, confidence, and behavioral responses to different types of spoofing attacks, including email, SMS, caller ID, and website spoofing, in Saudi Arabia. A survey was conducted to gather data from 1437 students at Saudi Electronic University, and it was analyzed using a quantitative research methodology and different statistical tests, such as Chi-square tests, Friedman tests, Kruskal–Wallis tests, correlation analysis, and regression models. The analysis results indicate that students exhibit a relatively high level of awareness. However, awareness and confidence vary across demographic groups, with significant differences associated with gender and age group. The results also reveal a significant gap between perceived confidence and detection ability in scenario-based assessments, highlighting that self-reported awareness does not necessarily translate into practical identification skills. The study emphasizes the importance of strengthening practical cybersecurity education, simulation-based training, and effective awareness delivery methods to improve students’ ability to recognize impersonation-based cyber threats in the Saudi educational sector.

Big Data and Cognitive Computing doi: 10.3390/bdcc10060169

Authors: Ismail Ifakir El Habib Nfaoui Abderrahim Zannou Asmaa Mourhir

Aspect-Based Sentiment Analysis explores sentiment trends related to specific opinion aspects and holds significant commercial potential for monitoring brand reputation, understanding customer satisfaction, and personalizing recommendations. However, traditional methods rely exclusively on textual input and often struggle when the target aspect is not mentioned in the sentence. Multimodal Aspect-Based Sentiment Analysis addresses this limitation by incorporating both textual and visual modalities to enable more comprehensive sentiment understanding. Despite advancements in deep learning and transformer-based architectures, existing models often suffer from suboptimal modality fusion and weak aspect grounding, limiting their classification accuracy. To overcome these challenges, we propose an Attention-Driven Feature Fusion (ADFF) approach based on a three-stage hierarchical attention mechanism. First, it only fuses text and image embeddings. Second, it incorporates aspect-level features. Third, a multi-head attention layer further enhances cross-modal dependencies. The resulting representation is passed to a Long Short-Term Memory (LSTM) classifier for sentiment polarity prediction. We evaluate our model on three benchmark datasets, namely Twitter-2015, Twitter-2017, and MASAD. The experimental results demonstrate that the proposed model substantially outperforms state-of-the-art multimodal and unimodal baselines, improves both accuracy and F1-score, achieving 82.55% accuracy and 81.05% F1-score on Twitter-2015, 77.07% accuracy and 77.15% F1-score on Twitter-2017, and up to 99.67% accuracy and F1-score in the Plant domain of MASAD, where we observe consistent improvements across all seven domains. These results highlight the effectiveness and scalability of the hierarchical attention-based fusion strategy for real-world aspect-based sentiment analysis tasks.

Big Data and Cognitive Computing doi: 10.3390/bdcc10060168

Authors: Alexandr V. Korchemnyi Alexey K. Kovalev Aleksandr I. Panov

The idea of disentangled representations is to reduce the data to a set of generative factors that produce it. Typically, such representations are vectors in latent space, where each coordinate corresponds to one of the generative factors. The object can then be modified by changing the value of a particular coordinate, but it is necessary to determine which coordinate corresponds to the desired generative factor—a difficult task if the vector representation has a high dimension. In this article, we propose ArSyD (Architecture for Symbolic Disentanglement), which represents each generative factor as a vector of the same dimension as the resulting representation. In ArSyD, the object representation is obtained as a superposition of the generative factor vector representations. We call such a representation a symbolic disentangled representation. We use the principles of Hyperdimensional Computing (also known as Vector Symbolic Architectures), where symbols are represented as hypervectors, allowing vector operations on them. Disentanglement is achieved by construction, no additional assumptions about the underlying distributions are made during training, and the model is only trained to reconstruct images in a weakly supervised manner. We study ArSyD on the dSprites and CLEVR datasets and provide a comprehensive analysis of the learned symbolic disentangled representations. ArSyD outperforms BetaVAE and FactorVAE baselines on CLEVR1 paired, achieving an FID of 93.72 compared to 129.68 and 115.61, respectively. It also achieves the best IOU value on dSprites paired, at 98.37, compared to 96.43 and 97.11 for the other baselines. We also propose new disentanglement metrics that allow comparison of methods using latent representations of different dimensions. ArSyD allows us to edit the object properties in a controlled and interpretable way, and the dimensionality of the object property representation coincides with the dimensionality of the object representation itself.

Big Data and Cognitive Computing doi: 10.3390/bdcc10060167

Authors: Salam Allawi Hussein Sándor R. Répás

The growing prevalence of encrypted network traffic has rendered traditional payload-based inspection ineffective, shifting attention toward flow-level statistical analysis combined with machine learning. At the same time, privacy regulations and distributed network architectures make centralised data collection increasingly impractical, motivating federated learning as a privacy-preserving alternative. Despite its promise, deploying federated learning for encrypted traffic classification in realistic environments remains challenging, particularly under heterogeneous client data distributions that arise when different network sites observe different subsets of services. This paper examines how server-side aggregation affects federated QUIC traffic classification under such heterogeneous conditions. We use a five-class Google QUIC dataset and represent each flow with eight statistical features derived from packet size and timing. We compare a centralised baseline with federated learning under three client partitions: mixed-label clients (C1), service-based single-class clients (C2), and hash-based semi-IID clients (C3). For each case, we evaluate four Flower aggregation strategies: FedAvg, FedAdam, FedAvgM, and FedYogi. Results show that client distribution has a greater impact on performance than the choice of aggregation strategy. Federated models match or closely approach centralised performance in C1 and C3, with accuracy up to 0.9969 and macro-AUC near 1.0. In C2, accuracy drops due to extreme label skew, but adaptive aggregation mitigates the effect. FedYogi achieves the best C2 accuracy of 0.9287, while FedAvgM attains the highest C2 macro-AUC of 0.9885. ROC curves and confusion matrices confirm that the choice of aggregation matters mainly under severe heterogeneity.

Big Data and Cognitive Computing doi: 10.3390/bdcc10060166

Authors: Alex Krempasky Erik Kajati Peter Papcun

Large Language Models (LLMs) offer opportunities for processing heterogeneous information streams relevant to energy-market decision-making, but their practical role in forecasting-oriented analytical workflows remains uncertain. This paper presents an exploratory feasibility study of LLM use across four energy-market tasks: geopolitical event monitoring for Dutch Title Transfer Facility (TTF) market context using Global Database of Events, Language, and Tone (GDELT)-based data, structured summarisation of commodity-intelligence articles, prompt-engineered solar-power and grid-load forecasting for Austria, and a short-horizon exploratory TTF price-estimation case. The study is positioned as a pilot investigation and hybrid workflow blueprint rather than as a statistically conclusive forecasting benchmark. A four-layer reference architecture was devised, including structured market data, semi-structured news intelligence, web-scraping concepts, and implemented Twitter/X and GDELT monitoring layers. The empirical cases indicate that LLMs are most useful for text-heavy reasoning, event-context integration, source triage, and structured interpretation. In the 20-article summarisation corpus, Gemini 1.5 Pro achieved higher commodity-direction accuracy than GPT-4, while GPT-4 showed stronger output-format stability. In selected solar case checks, OpenAI models produced plausible generation curves close to the Fraunhofer ISE Energy Charts reference, while Energy Charts remained more accurate for aggregate load estimation in the available benchmark comparison. The two-day TTF experiment illustrated that LLMs can incorporate qualitative geopolitical context into short-horizon reasoning, but it did not establish reliable price-forecasting capability. The Twitter/X monitoring layer is retained as a documented negative pathway, showing the limitations of informal social-media scraping for reproducible market intelligence.

Big Data and Cognitive Computing doi: 10.3390/bdcc10050165

Authors: Virginia Puyana-Romero Angela María Díaz-Márquez Christiam Santiago Garzón-Pico Giuseppe Ciaburro

The change in learning methods to online/distance learning, catalyzed by recent health pandemics/social distancing requirements, has significantly changed how teaching occurs and what students experience in their learning spaces in regard to interference. New forms of interference exist, and they are related to the domestic setting of the student’s life. This study examined how factors of domestic life influence what students find in regard to interference in their online learning spaces through a Likert-scale defined answer process to a 29-question predictor variable inventory that also includes two outcome variables that address the amount of acoustic interference experienced in learning spaces. Moreover, through regression models and various applications of machine learning science, this research aims to reveal crucial indicators that influence student experiences regarding disturbances. In this respect, these findings highlight crucial roles that housing density and internal interactive actions within residential contexts have on disturbances. Furthermore, this research reveals critical understandings of perceptual inequalities present within distance learning student populations and indicates significant cultural and social consequences related to digital technologies. This is crucial, understood within foundational perspectives that are necessary to address psychosocial challenges and human–building interaction present within distance learning science and policies aimed at reducing noise.

Big Data and Cognitive Computing doi: 10.3390/bdcc10050164

Authors: Lee Naish

Identifying similarities in data is fundamental to discovery in science. Measuring or ranking similarity is a key way of reducing the dimensionality of data, is at the heart of many data intensive algorithms and can also be used directly for some applications. This paper extends our understanding of a relatively simple similarity problem. Our primary application is spectral-based fault localisation (SBFL), in which a computer program is run with a large number of test cases and data is collected on which statements are executed in each test case. For each statement, the set of test cases in which it is executed is compared to the set of test cases that failed, and this is used to rank the statements to help locate bugs, an instance of what we call the similarity to a single set (STASS) problem. This paper is primarily theoretical but some contributions are validated with SBFL experiments. Set similarity is equivalent to similarity of binary vectors or two-by-two contingency tables. The problem is also equivalent to converting two-dimensional data with a “partial order”, such as points on a rectangular grid, to a one-dimensional total order. Even when the raw data is not binary, we are often interested in comparing binary classifiers for the data, such as diagnostic tests, and comparing binary classifiers is an instance of the STASS problem. More than a hundred set similarity measures have been proposed in the literature and hundreds of thousands have been evaluated for SBFL, but there is very little understanding of how best to choose a similarity measure for a given domain. This work discusses numerous properties and forms of symmetry that similarity measures can have. It refines previously identified properties so they are no longer incompatible, identifies new forms of symmetry, defines ordering relations over similarity measures, and proposes a new statistic that can be used to help choose a good similarity measure for a given domain.

Big Data and Cognitive Computing doi: 10.3390/bdcc10050163

Authors: Borja Arroyo Galende Patricia A. Apellániz Alejandro Almodóvar Silvia Uribe Federico Álvarez Juan Parras

We study Membership Inference Attack (MIA) risk in Federated Learning through a two-stage lens that separates (i) whether a target client’s contribution is detectable after aggregation and system noise (Stage I: Signal Survival) from (ii) whether a surviving contribution induces a generative membership score change attributable to the target’s private data (Stage II: Signal Attribution). Stage I models aggregation as a target–background decomposition and shows that detectability hinges on target–background alignment, which can induce cancellation. Stage II connects the surviving target component to a generative MIA score via a local path representation and Lipschitz/smoothness bounds, avoiding architecture-specific assumptions. Our analysis reveals that the leading attribution term is governed by the alignment between the target update and the score geometry of the target data at an appropriate baseline. We validate the theoretical bounds and illustrate risk trajectories across several scenarios.

Big Data and Cognitive Computing doi: 10.3390/bdcc10050162

Authors: Andreas Polyvios Delladetsimas Elias Iosif Stamatis Papangelou George Giaglis

This article examines blockchain as an enabling technological component for data management tasks that are independent of currency-related functionality, a less-discussed aspect of a technology commonly associated with cryptocurrencies and decentralized finance (DeFi). Drawing on empirical findings from the DIGI4ECO project as a case study, we present a structured literature review and cross-domain analysis of blockchain-based data management systems (BDMSs), examine a representative permissioned BDMS implementation, and synthesize practical design guidelines and implementation insights for BDMS development. This perspective is motivated by core blockchain properties such as immutability and transparency, as well as by the observation that existing resources for BDMS development, including methods, tools, and best practices, remain fragmented and less developed than those available for more mature technologies.

Big Data and Cognitive Computing doi: 10.3390/bdcc10050161

Authors: Fumin Zou Lei Zou Feng Guo Xunhuang Wang Jianqing Weng Tao Fang Haocai Jiang Xueming Wu

This study introduces the Quantum-inspired Pretrained Feature Embedding (ImprovedQPFE) model, a framework for dialogue sentiment classification. ImprovedQPFE integrates phase-pretrained complex embeddings, a bidirectional complex-valued GRU, a quantum-inspired attention mechanism, and supervised contrastive learning within a Transformer-based architecture, aiming to enhance feature discriminability under class imbalance. We evaluate ImprovedQPFE on the RECCON-DD and RECCON-IEM benchmarks under a unified and reproducible protocol, including standardized preprocessing and fixed data splits. To ensure reproducibility, all experiments were conducted using a fixed random seed of 42. The reported results are based on this single fixed-seed setting rather than averages over multiple repeated runs. The empirical results show that ImprovedQPFE achieves competitive performance and outperforms the compared baselines under the adopted experimental protocol. On the RECCON-DD dataset, ImprovedQPFE improves Macro-F1 from 80.08% to 83.75% compared with a strong non-quantum Transformer-based baseline equipped with contrastive learning. It also improves Pos-F1 while maintaining high performance for negative classes. On RECCON-IEM, ImprovedQPFE attains a leading Macro-F1 of 95.39% among the compared methods. These findings, together with an ablation analysis, support the effectiveness of the proposed quantum-inspired representation paradigm and its architectural components. However, further statistical validation with multiple repeated runs, standard deviations, confidence intervals, and significance testing remains an important direction for future work.

Big Data and Cognitive Computing doi: 10.3390/bdcc10050160

Authors: Nithya Nedungadi Sriram Sankaran Krishnashree Achuthan

The evolution from the Internet of Things (IoT) to the Internet of Vehicles (IoV) has expanded intelligent connectivity across embedded systems while increasing cybersecurity risks arising from large scale data exchange and device heterogeneity. As IoV environments become more dynamic and safety critical, centralized Intrusion Detection Systems (IDSs) face constraints related to latency, privacy exposure, and bandwidth overhead. These limitations motivate a transition to edge-enabled IoV architectures, where localized vehicular and anchor nodes supported by edge servers enable decentralized processing, enhanced privacy, and reduced communication load. To address these operational challenges, this paper proposes FedX (Federated Explainable Ensemble Intrusion Detection System), a privacy-preserving and explainable federated ensemble IDS that integrates XGBoost and LightGBM models across resource-constrained edge vehicles and roadside units (RSUs) to enable collaborative, low-latency anomaly detection without sharing raw data. By applying adaptive weighting based on model confidence and resource availability, FedX enhances robustness and efficiency while enabling explainable decisions via SHAP and LIME analysis, which highlights reliance on key features (flow duration, speed, RPM) for high-confidence (>97%) intrusion alerts grounded in domain-specific behavior. Privacy is further enforced through Gaussian differential privacy and secure aggregation to mitigate inference and inversion attacks. Experiments on the CICIoV2024 dataset show that FedX achieves 99.1% accuracy, outperforming existing federated ensemble IDS models by up to 2.1%. The system reduces communication overhead by 17% relative to full synchronization through adaptive weighted transmission and secure aggregation. It maintains negligible accuracy loss (<1.5%) under a strong privacy budget (ϵ = 1.1). The deployment of proposed IDS on Raspberry Pi 4 underscores its efficacy for edge computing. Experimental results indicate that adaptive weighting yields a 1.8% performance increase, while resource profiling shows 45% lower CPU utilization and over 50% lower power consumption compared with centralized baselines. The findings demonstrate that FedX, combined with explainable AI enables trustworthy, interpretable, and energy-efficient intrusion detection for secure next-generation Edge-enabled IoV networks.

Big Data and Cognitive Computing doi: 10.3390/bdcc10050159

Authors: Md Enamul Kabir Shakhawat H. Tanim Deanna D. Sellnow Geneva Lei P. Luteria Lior Rennert

Vaccine hesitancy—which can be defined as a delay in acceptance or the refusal to get vaccinated—has substantially increased over the past decade. This study introduces a computational and qualitative approach designed to efficiently classify stance and uncover narratives in social media discourse without relying on extensive manual annotation. Using 298,356 COVID-19 vaccine-related X posts geolocated to South Carolina (June 2021–May 2022), zero-shot and few-shot learning with instruction-tuned large language models (Mistral-7B, Meta-Llama-3.1, and DeepSeek-7B) was applied for stance detection while Latent Dirichlet Allocation (LDA) was used for topic modeling. The topic modeling identified five dominant themes in vaccine hesitant conversations: skepticism of vaccine efficacy, comparative framing, scientific justification, disapproval of regulations, and distrust. Temporal analysis revealed that skepticism peaked during late 2021, coinciding with booster campaigns and mandate debates. These findings suggest that vaccine hesitancy is influenced through complex rhetorical strategies rather than misinformation alone. These underlying narratives often frame skepticism as rational and evidence-based, using scientific language and statistical reasoning to challenge the effectiveness of vaccines.

Big Data and Cognitive Computing doi: 10.3390/bdcc10050158

Authors: Ahmed El-Kosairy Heba Kamal Aslan

Many blockchain-based cyber threat intelligence (CTI) sharing systems emphasize immutability and auditability, but often treat CTI submissions as ordinary blockchain transactions without explicitly separating content validation from publication anchoring. This paper presents CTIB, a proof-of-concept hybrid Proof-of-Stake (PoS) and Proof-of-Work (PoW) framework for CTI publication. CTIB uses a sequential workflow in which a PoS committee first evaluates CTI submissions, and an accepted feed hash is then anchored through a PoW step to provide verifiable temporal binding. The prototype is evaluated in a controlled local Hardhat environment; therefore, the results should be interpreted as prototype-level feasibility evidence rather than production-scale deployment results. CTI content is represented using STIX 2.1, canonicalized, and hashed using SHA-256; only integrity-critical evidence is stored on-chain, while full CTI content remains off-chain. Experimental results demonstrate prototype-level feasibility, with measured throughput, latency, and success rate metrics under different PoW difficulty profiles. Across ten independent local runs, CTIB achieved an average throughput between 141.13 and 166.14 feeds/min, average p50 latency between 326.18 and 403.09 ms, and average p95 latency between 553.22 and 700.82 ms under the tested difficulty profiles. Security analysis uses analytical modeling, committee capture probability, and Monte Carlo simulation to evaluate majority-attack feasibility under stated assumptions. The results indicate that sequential compromise of both validation and anchoring layers increases the cost of coordinated manipulation.

Big Data and Cognitive Computing doi: 10.3390/bdcc10050157

Authors: Yonghong Chen Guoqi Zhang Wanzhi Wen Shibing Zhang

Speech produced in helium–oxygen (heliox) environments in deep saturation diving exhibits pronounced spectral shifts and temporal distortions, which severely degrade automatic speech recognition (ASR) systems trained on normal-air corpora. Existing studies often adopt a restoration-then-recognition paradigm by training waveform mapping networks on paired heliox/air recordings. However, in realistic low-resource data collection, paired recordings are typically obtained by independent re-reading and are therefore not strictly time-aligned, which makes regression-style restoration more sensitive to pairing errors and increases the risk of front-end distortions. This paper proposes a robust recognition framework for heliox speech, termed PRL-DAS (Physics-informed Resampling and LoRA with Duration-Adaptive Speed). The framework consists of a physics-inspired linear resampling warm start (PhysSpeed), parameter-efficient Low-Rank Adaptation (LoRA), and duration-adaptive speed (DAS) inference enhancement. Specifically, we first apply physics-motivated linear resampling as a coarse warm start, and then perform mixed-domain LoRA fine-tuning of a Whisper foundation model to absorb residual non-linear differences. On a corpus of 1048 paired Chinese heliox utterances under leave-one-speaker-out (LOSO) evaluation, using Whisper-Medium as the base model, PhysSpeed followed by mixed-domain LoRA reduces the overall character error rate (CER) from 49.33% with PhysSpeed preprocessing only to 25.79%, while also improving performance on the normal domain. Furthermore, the full PRL-DAS framework applies Soft-DAS, a lightweight smooth schedule motivated by duration-dependent variation in the optimal resampling factor, and further reduces the overall CER to 24.37% without additional training cost.

Big Data and Cognitive Computing doi: 10.3390/bdcc10050156

Authors: Nikita Stepanov Anastasiya Radaeva Peter Panfilov Alexander Suleykin Valery Pyatetsky

University admissions services must answer large volumes of applicant questions that differ substantially in complexity, ranging from repetitive FAQ-type requests to multi-step questions involving programs, entrance exams, admission rules, passing scores, and temporal comparisons. Ungrounded large language model responses are risky in this domain because answers must be factually correct, source-based, and consistent with official institutional data. This paper presents a three-tier hybrid architecture for an admissions dialogue assistant that combines deterministic FAQ matching, hybrid retrieval-augmented generation, and graph-grounded retrieval for complex queries. The first tier, Hash-FAQ, returns verified answers for frequent intents using normalized keys, hash-based lookup, near-duplicate fingerprinting, and semantic similarity checks. The second tier applies hybrid RAG based on BM25 retrieval, vector search, rank fusion, and optional cross-encoder reranking. The third tier uses GraphRAG to extract a constrained k-hop subgraph from a Neo4j knowledge graph built from relational admissions data and document-derived facts. All tiers are synchronized through a versioned indexing pipeline with shadow collections and atomic switching across lexical, vector, FAQ, relational, and graph stores. The system was evaluated using real admissions-campaign traffic and a labeled subset of applicant queries. Tier 1 resolved 68.7% of requests with low latency, while the GraphRAG branch improved factual accuracy with attribution on multi-step queries from 0.55 to 0.91 compared with the non-graph baseline. The main contribution of the study is a production-oriented, cost-aware retrieval-and-generation architecture that links tiered routing, synchronized knowledge publication, source attribution, and operational evaluation for applicant-facing institutional dialogue systems.

Big Data and Cognitive Computing doi: 10.3390/bdcc10050155

Authors: Josip Tomo Licardo Nikola Tanković Ivan Osman Ivan Lorencin Sandi Baressi Šegota

Large Language Models (LLMs) offer state-of-the-art performance in natural language understanding and generation tasks. However, the deployment of leading commercial models for specialized tasks, such as e-commerce, is often hindered by high computational costs, latency, and operational expenses. This paper investigates the viability of smaller, open-weight models as a resource-efficient alternative. We present a methodology for optimizing a one-billion-parameter Llama 3.2 model for multilingual e-commerce intent recognition. The model was fine-tuned using Quantized Low-Rank Adaptation (QLoRA) on a synthetically generated dataset designed to mimic real-world user queries. Subsequently, we applied post-training quantization techniques, creating GPU-optimized (GPTQ) and CPU-optimized (GGUF) versions. Our results demonstrate that the specialized 1B model achieves 98.8% accuracy, approaching the performance of the significantly larger GPT-4.1 model. A detailed performance analysis revealed critical, hardware-dependent trade-offs: while 4-bit GPTQ reduced VRAM usage by 41%, it paradoxically slowed inference by 82% on an older GPU architecture (NVIDIA T4) due to dequantization overhead. Conversely, GGUF formats on a CPU achieved a speedup of up to 4.3× in inference throughput and up to a 72% reduction in RAM consumption compared to the FP16 baseline. We conclude that small, properly optimized open-weight models are not just a viable but a more suitable alternative for domain-specific applications, offering state-of-the-art accuracy at a fraction of the computational cost.

Big Data and Cognitive Computing doi: 10.3390/bdcc10050154

Authors: Miljenko Švarcmajer Mirko Kohler Zdravko Krpić Ivica Lukić

Modern data-driven distributed systems increasingly rely on blockchain technologies to ensure trust, transparency, and decentralized coordination. However, the rapid proliferation of consensus mechanisms has created a complex design space, making the selection of an appropriate protocol a non-trivial architectural and decision-making challenge. Different consensus mechanisms rely on distinct security resources, validator admission models, and agreement architectures, leading to diverse trade-offs between scalability, decentralization, performance, and governance. Existing studies primarily focus on classification or performance comparison of consensus mechanisms, while the problem of systematic, requirement-driven selection remains insufficiently addressed. In particular, there is a lack of structured approaches that integrate multiple system requirements into a unified decision framework suitable for real-world environments. To address this gap, this paper proposes a consensus-driven, layered framework for blockchain consensus mechanism selection, formulated as a multi-criteria decision problem. The framework organizes the consensus design space across key architectural dimensions and analyzes 32 consensus mechanisms, enabling systematic comparison and supporting data-driven decision-making. The approach is further demonstrated through five representative use-case scenarios, showing its applicability in optimizing distributed system design.

Big Data and Cognitive Computing doi: 10.3390/bdcc10050153

Authors: Yong-Wei Zhang Ming-Yang Zhu Wen-Kai Xia Xin-Yang Zhang Jin-Di Liu

Learning path optimization is crucial in intelligent educational systems, with the core challenge of efficient multi-objective sequential decision-making under complex prerequisite constraints. To address the poor generalization of existing methods relying on fixed operator scheduling or handcrafted heuristics, this paper proposes a hyper-heuristic framework based on Dueling Deep Q-Network (Dueling DQN-HH), formulating operator selection as a sequential decision-making process for dynamic adaptive scheduling of low-level operators. The framework adopts priority-based encoding to unify learning path representation (decoupling the hyper-heuristic layer from the problem domain) and designs a composite reward mechanism integrating reward shaping, exploration incentives, and computational cost awareness to balance solution quality and efficiency. Additionally, it employs a dueling network architecture with prioritized experience replay to enhance policy learning stability. Experimental results show the proposed method outperforms representative baseline algorithms in solution quality, convergence stability, and computational efficiency. The framework demonstrates superior performance across multiple objectives, particularly in minimizing the total learning time (Ftime), as validated on two heterogeneous datasets: MOOCCube (Computer Science) and PsyDataset (Psychology). Further ablation studies and operator evolution analyses verify its adaptive scheduling capability under different objectives and knowledge graph structures, demonstrating strong objective independence and cross-dataset generalization.

Big Data and Cognitive Computing doi: 10.3390/bdcc10050152

Authors: Michał Gostkowski Tomasz Ząbkowski Krzysztof Gajowniczek

Effective peak load management enables utilities to mitigate increased electricity demand and optimize the use of available resources during periods of maximum consumption. Accurate forecasting of the peak load is essential for ensuring the reliability, efficiency, and resilience of contemporary power systems. In this study, commercial customer-level data were employed to identify electricity peak demand within the Polish power system, drawing upon historical records of both energy consumption and meteorological variables. Departing from conventional time series forecasting approaches, the problem was intentionally reformulated as a pattern recognition task. Three classification techniques were systematically evaluated to identify individual customers’ peak load events, thereby offering a basis for demand-side management strategies and incentive mechanisms aimed at flattening load profiles and improving grid stability. The proposed approach demonstrates how data-driven analytics can support utilities in extracting actionable knowledge from large-scale energy datasets and enabling proactive demand response programs. Empirical results indicate that the proposed methods are capable of predicting up to 90% of electricity peak occurrences, with a forecasting horizon of 24 h leading to significant shifts in the load curve.

Big Data and Cognitive Computing doi: 10.3390/bdcc10050151

Authors: Roland Tanácsi András Micsik

Background: The Hungarian Science Bibliography applies the OECD Frascati Fields of Science and Technology taxonomy for subject classification; however, approximately 80% of its records lack assigned categories. Automated large-scale classification could support retrospective completion and improve the quality of bibliographic data. Methods: We evaluated multiple artificial intelligence approaches to classifying publications into level 4 Frascati categories using only titles and keywords. Training datasets were compiled from bibliographic records and subjected to heuristic and large-language-model-based filtering to reduce noise and ambiguity. The approaches tested included statistical methods, classical machine learning classifiers, fine-tuned SciBERT models, zero-shot prompting with large language models, and a Mixture-of-Experts architecture. Results: Data quality had a stronger impact on performance than model complexity. Large-language-model-based filtering substantially improved classification results. The best-performing model, a Support Vector Classifier, achieved a weighted F1 score of 0.83, which is an outstanding result relative to state-of-the-art approaches from the literature. Conclusions: Our findings contribute new insights into classification research and may assist others in selecting appropriate solutions for real-world, large-scale bibliographic classification tasks.

Big Data and Cognitive Computing doi: 10.3390/bdcc10050150

Authors: Hao Nguyen Thi Loc Nguyen The Huu Dang Quoc

This paper introduces the ETMS-RCPSP (Earliness–Tardiness Multi-Skill Resource-Constrained Scheduling Problem)—a novel problem derived from the MS-RCPSP by adding constraints on project completion time or actual production contracts. The goal of the new problem is to control the project completion time as closely as possible to reality—this differs from the original MS-RCPSP, which aimed to minimize project execution time. The objective of the problem is of greater practical significance in ensuring project completion on schedule while also addressing related issues, such as the ability to receive finished products on time as stipulated in the contract. The ETMS-RCPSP is an NP-hard problem whose result can be used for resource allocation in project execution or for resource arrangement in production lines to fulfill economic contracts. To address the ETMS-RCPSP, the paper proposes a new evolutionary algorithm, BDERL (Balanced Differential Evolution with Reinforcement Learning), that combines differential evolution with a problem-specific decoding mechanism and an adaptive parameter control strategy based on reinforcement learning (Q-learning). The proposed algorithm is evaluated on benchmark instances derived from the iMOPSE dataset and the TNG company dataset—a real-world dataset from manufacturing and contract-driven environments. Experimental results demonstrate that the approach consistently reduces total production costs compared to baseline heuristics while maintaining competitive computational efficiency. The findings underscore the efficacy of adaptive hybrid optimization techniques in solving intricate production scheduling problems characterized by limited resources and varied skill competencies.

Big Data and Cognitive Computing doi: 10.3390/bdcc10050149

Authors: Shaofu Lin Yumiao Chang Jianhui Chen Lianfang Ma

Accurate kidney tumor segmentation from abdominal CT is essential for quantitative assessment and treatment planning. However, indistinct tumor boundaries and substantial inter-patient shape variability render traditional hand-crafted feature-based methods unreliable for precise delineation. Although deep learning has advanced this task, these methods still struggle with multi-scale tumor characteristics, complex morphological variations, and background noise in medical images. To address these challenges, we propose MDA-Net, an end-to-end segmentation method based on enhanced multi-scale feature extraction and attention refinement. Specifically, we introduce a Multi-Scale Feature Extraction (MSFE) module into encoder–decoder skip connections to aggregate dilated features across multiple receptive fields and learn branch-wise weights for adaptive refinement and fusion, thereby enhancing boundary details and semantic cues to reduce tumor-tissue ambiguity. At the bottleneck, a Deformable Pyramid Feature Refinement (DPFR) module combines deformable sampling with pyramid contextual modeling, thereby improving adaptability to variations in tumor shape and scale while preserving feature resolution. Moreover, a Channel and Spatial Attention (CASA) module is embedded in the decoder to suppress background interference and enhance boundary-sensitive structures during upsampling via coordinated channel and spatial reweighting, thereby improving the reconstruction of fine-grained tumor morphology and contours. Experiments on both KiTS19 and KiTS21 show that MDA-Net consistently improves tumor boundary delineation, lesion localization, and mask reconstruction, demonstrating stronger robustness and cross-dataset generalizability than representative baseline methods. Ablation studies further confirm the complementary effects of MSFE, DPFR, and CASA. In addition, Grad-CAM visualizations improve the clinical transparency and interpretability of the model. Overall, this method advances deep learning for medical image analysis and supports precise diagnosis and treatment of renal tumors.

Big Data and Cognitive Computing doi: 10.3390/bdcc10050148

Authors: Serhii Vladov Oksana Mulesa Maryana Marusinets Tiberiy Chegi Victoria Vysotska Anton Kazakov Iryna Kirieieva Maksym Korniienko Tetiana Morhunova

The article develops a hybrid multimodal neural network for the automatic prevention of personal digital rights violations, focusing on improving security through anomaly detection and ensuring data confidentiality. The main aim is to integrate several innovative methods, such as federated learning, gating, latent competitive learning, and a variational autoencoder, to improve violation detection accuracy. The key contribution is the development of a training mixture that combines a probabilistic anomaly detector and an autoencoder reconstruction signal, which allows for effective detection of typical incidents and hidden anomalies. The experimental evaluation results showed high-performance indicators, with ROC-AUC at 0.96 and accuracy at 0.94, confirming the system’s effectiveness on anonymized data. The results obtained have a significant practical contribution, as they can be integrated into national information security systems, including SOC and forensic reports, which will ensure a higher level of personal data protection and reduce privacy breach risks. The scope of the proposed system simultaneously covers cybersecurity, personal data protection, national security, SOC systems, and forensic analysis.

Big Data and Cognitive Computing doi: 10.3390/bdcc10050147

Authors: Yihua Zhu Chao Luo Ke Wu Jiawei Yu Binjiang Hu Lei Huang Bitao Xiao

Enhancing the accuracy of short-term wind power forecasting helps mitigate the adverse impacts of prediction errors on grid dispatch. Wind power exhibits a significantly nonlinear dependence on multiple influencing factors. However, existing methods struggle to effectively resolve multi-dimensional feature redundancy and multi-scale non-stationary evolutionary characteristics inherent in far-offshore wind power forecasting tasks. This leads to bottlenecks such as insufficient feature discriminability and temporal dependency focus shift under complex marine environments, ultimately limiting further improvements in prediction accuracy. To address these challenges, this paper proposes a federated learning-based adaptive multi-head attention model for wind power forecasting (Fed-AMHA). The proposed framework operates as follows: First, each wind farm client utilizes a Bidirectional Long Short-Term Memory (BiLSTM) network to model input sequences bidirectionally, capturing long-term temporal dependencies. Subsequently, linear projection and parallel one-dimensional convolution operations are introduced to mine multi-scale local temporal features from each time step and its neighborhood. Building upon this, channel attention and multi-head temporal feature attention mechanisms are stacked. The model adaptively adjusts the weights of different time slices and feature channels by learning the importance of each channel to the forecasting task. The central server then aggregates the model parameters uploaded by the clients via averaging, enabling cross-site collaborative training without directly sharing raw data. Simulation results based on public datasets and actual wind farm data under various short-term forecasting scenarios demonstrate that the proposed model consistently achieves lower prediction errors and superior stability compared to existing forecasting models under identical settings.

Big Data and Cognitive Computing doi: 10.3390/bdcc10050146

Authors: Elia Morgulev Ofer H. Azar

Big-data analytics are increasingly used in scouting and talent identification, with machine learning (ML) tools applied to evaluate and predict player performance based on match statistics, video tracking, physical and anthropometric tests, psychological assessments, social media data, and qualitative scouting reports. Advances in computer vision, together with the emergence of affordable automated broadcasting and data collection systems, have extended the deployment of ML-driven scouting from professional to youth sport. The use of algorithms in educational, employment, and healthcare settings has been shown to introduce biases and discrimination while wrongly assuming accuracy and objectivity because the decisions are made automatically and quantitatively. In this respect, we briefly describe the development of data-driven performance analysis and how ML-based technologies are currently applied for early screening and comparison of large player populations. Based on a narrative overview of the literature, we draw on evidence from education, employment, and healthcare to identify risks that may also emerge in ML-driven player evaluation, including algorithmic bias, non-representative training data, privacy concerns, and the persistence of model-based labels over time, especially in youth sport. Our main contribution is translating these threats into governance principles and operational safeguards for responsible use of AI in scouting and talent identification.

Big Data and Cognitive Computing doi: 10.3390/bdcc10050145

Authors: Rodrigo Gabriel Salazar Alva Matías Núñez Cristian López Del Alamo Javier Martín Arista

Preserving ancient languages is essential for understanding the cultural and linguistic heritage of humanity. Old English, however, remains critically under-resourced, which limits its accessibility to modern natural language processing (NLP) techniques. We present a scalable framework that uses advanced large language models (LLMs) to generate high-quality Old English texts to address this gap. In this study, we specifically employ state-of-the-art models, including Llama-3.1-8B and Mistral-7B, as our foundation models, which are then adapted to the unique characteristics of Old English. Our approach combines parameter-efficient fine-tuning (Low-Rank Adaptation (LoRA)), data augmentation via back-translation, and a dual-agent pipeline that separates content generation (in English) and translation (into Old English). Evaluation with automated metrics (BLEU, METEOR, and CHRF) shows improvements over baseline models, with BLEU scores increasing from 26 to over 65 for English-to-Old English translation. Expert human assessment confirms high grammatical accuracy and stylistic fidelity in the generated texts, with average scores of 9.0/10 for inflection and word order, 9.1/10 for lexical authenticity, and 7.8 for semantic coherence. These results demonstrate that the framework can reliably expand limited historical corpora while maintaining linguistic integrity, with immediate practical applications in digital humanities research, computational philology, and the development of educational resources for Old English study. Beyond expanding the Old English corpus, our method offers a practical blueprint for revitalizing other endangered languages, thus linking AI innovation with the goals of cultural preservation.

Big Data and Cognitive Computing doi: 10.3390/bdcc10050144

Authors: Ismail Jabri Zine Eddine Louriga Aziza El Ouaazizi Abdelaziz Ahaitouf

Sarcasm detection remains a challenging task in natural language processing because sarcastic expressions often convey meanings that contradict their literal wording. Although transformer-based encoders such as RoBERTa capture contextual semantics effectively, sparse linguistic signals common in sarcastic user-generated text, such as exaggerated punctuation, elongated words, capitalization, and sentiment contrast, may not always remain explicitly accessible in the final sentence representation. To address this limitation, we propose HYSARD, a hybrid feature-fusion model that combines RoBERTa-based sentence embeddings with complementary linguistic features, including sentiment polarity, stylistic markers, syntactic patterns, and TF-IDF lexical cues. The resulting feature space is refined through Random Forest-based feature selection to reduce redundancy and improve robustness, while SMOTE mitigates class imbalance during training. We evaluate HYSARD on the SemEval-2022 iSarcasmEval dataset and the balanced Main and Political subsets of SARC 2.0. Results show strong and consistent performance across datasets, with an F1-score of 0.80 on iSarcasmEval, while held-out test-set error analysis further highlights strong class-wise discrimination. The ablation study further confirms that combining contextual embeddings with explicit linguistic cues improves sarcasm detection over reduced feature configurations. These findings show that hybrid feature fusion remains an effective and practical strategy for sarcasm detection in noisy social media text.

Big Data and Cognitive Computing doi: 10.3390/bdcc10050143

Authors: Itai Himelboim Mudit Baid

This manuscript develops and demonstrates a practical framework for evaluating automated classifiers used in communication research, using harmful language detection as an illustrative case. We combine (a) a structured review of documentation practices for 27 publicly available classifiers and their associated annotation processes with (b) a cross-dataset evaluation that re-tests each model beyond its original training context. Across 27 datasets, we extract and compare reporting on construct definitions, annotator instructions, and inter-annotator agreement, and we quantify generalization by applying each model to multiple out-of-domain test sets. We also benchmark a contemporary large language model (GPT-5) under a consistent prompting protocol to illustrate how LLM-based classification compares to fine-tuned classifiers. Results show that documentation is uneven and often insufficient for theory-driven measurement, inter-annotator agreement varies widely across datasets, and cross-dataset performance frequently drops substantially relative to within-dataset evaluations. Building on these findings and existing validation guidance, we provide a reusable checklist and decision flow to help researchers select, justify, and report classifier-based measures in ways that support transparency and cumulative science. Recommendations for researchers, reviewers, and journal editors stress aligning model selection with standards of validity, reliability, and transparency.

Big Data and Cognitive Computing doi: 10.3390/bdcc10050142

Authors: Emmanuel Chibuikem Nnadozie Susana Merino-Caviedes Daniel A. de Luis-Román Marcos Martín-Fernández Carlos Alberola-López

High-performance medical segmentation models are often benchmarked on high-end GPUs. Such benchmarks do not provide useful performance insights for point-of-care low-end devices. This work, firstly, posits that to achieve improved clinical adoption of AI-powered segmentation models, especially in reduced manpower settings like rural hospitals, we need benchmarks that provide actionable insights on the degree to which high-performance models address five deployment constraints viz: resource-effectiveness for low-end computing devices, clinically acceptable accuracy, clinically compatible execution times, localization of user data, and user-based finetuning. In this work, five state-of-the-art foundation segmentation models and one target-specific model were systematically evaluated on three multi-organ medical datasets. Furthermore, the best-ranking foundation model and target-specific model were benchmarked on three low-end devices. Our findings show that lightweight foundation models provided the best performance trade-off and are easily user-fine-tuned on custom datasets. Target-specific models provide high accuracy out-of-the-box, but may require significant optimisation to deliver comparably fast execution times and user-based finetuning on low-end devices. The methods and results from this research provide actionable insights on high-performance medical segmentation models for low-end computing devices, as a necessary step towards improved adoption in resource-limited clinical settings.

Big Data and Cognitive Computing doi: 10.3390/bdcc10050141

Authors: Fan Dong Wenjing Guo Jie Liu Ann Varghese Weida Tong Tucker A. Patterson Huixiao Hong

Social media platforms host abundant and timely descriptions of medication experiences that can complement traditional pharmacovigilance systems. Yet the linguistic informality of these data presents a major challenge for mapping adverse drug event (ADE) expressions to standardized medical terminology. In this study, we developed BERT-based language models to classify ADE mentions from social media into MedDRA System Organ Classes (SOCs). Using the SMM4H and CADEC corpora, as well as their combination, we performed 20 iterations of 20% holdout validation for 3-, 6-, 22-, and 25-SOC classification tasks with a selected fixed training configuration (learning rate, batch size, and training epochs) based on training-loss convergence. The models achieved accuracies ranging from 75% to 94%, demonstrating strong performance for SOC-level classification of noisy and informal ADE expressions under the evaluated settings. These results are based on a controlled mention-level evaluation using deduplicated adverse drug event strings and do not establish document-level or real-world deployment generalization. This work provides a systematic evaluation of BERT-based models for SOC-level classification of ADEs and demonstrates consistent performance within the evaluated datasets and label granularities. While direct comparison with prior studies is limited by differences in datasets and evaluation protocols, the results demonstrate that transformer-based models can effectively classify ADEs into SOCs. These findings support the use of transformer-based normalization for SOC-level aggregation of user-reported adverse events and their integration into large-scale social media pharmacovigilance pipelines as a downstream component under controlled conditions.

Big Data and Cognitive Computing doi: 10.3390/bdcc10050140

Authors: Yair Wiseman

The AVIF (AV1 Image File Format) is a modern, royalty-free image format that leverages the AV1 video codec for superior compression efficiency, supporting both lossy and lossless modes. Its entropy encoding relies on a multi-symbol context-adaptive arithmetic coder (range coding with adaptive cumulative distribution functions (CDFs)), which is effective for general imagery but may not optimally exploit the repetitive structures common in Geographic Information System (GIS) maps/data. This paper proposes replacing AVIF’s entropy encoder with the Burrows–Wheeler Transform (BWT), a reversible preprocessing algorithm that rearranges data to create runs of similar symbols, enhancing subsequent compression. We detail the technical steps for modification, drawing from AV1’s open-source implementation, and explain why BWT is advantageous for GIS raster maps/data, which often feature large uniform areas, limited color palettes, and spatial redundancies. Empirical evidence from related studies on BWT-based image compression shows improvements in lossless scenarios, potentially considerably reducing file sizes over standard methods while preserving data integrity critical for geospatial analysis. This swap could improve storage, transmission, and processing efficiency in GIS applications, such as remote sensing and cartography. The discussion includes challenges like computational overhead and compatibility, with recommendations for implementations. The resulting BWT-AVIF hybrid produces a non-standard AV1 bit-stream that is not compliant with the AV1 or AVIF specifications and therefore requires custom decoders. It is presented here as a research prototype for GIS-specific compression rather than a compliant AVIF extension.

Big Data and Cognitive Computing doi: 10.3390/bdcc10050139

Authors: Sunisa Kunarak

Next-generation heterogeneous wireless networks (HetNets) comprising LTE macro-cells, 5G New Radio (NR) small cells, and WiFi 6 access points aim to provide seamless connectivity under diverse mobility scenarios. However, vertical handover (VHO) remains a performance bottleneck because of the highly variable radio environments, dynamic user mobility, stringent quality of service (QoS) requirements, and the coexistence of multi-tier access technologies. Existing handover approaches based on deep learning and deep reinforcement learning (DRL) suffer from limitations: deep learning models lack decision-making capabilities, whereas DRL models, particularly deep Q-network (DQN)-based policies, face Q-value overestimation and unstable convergence. To overcome these limitations, this paper introduces a Hybrid deep double-Q networks (DDQN)–bidirectional long short-term memory (Bi-LSTM) Framework that integrates bi-directional mobility prediction and DRL-based adaptive decision-making. The Bi-LSTM module captures forward and backward temporal dependencies and predicts future Received Signal Strength (RSS) trajectories, mobility dynamics, and cell-edge transitions. The DDQN module stabilizes the action value estimation, mitigates overestimation bias, and enables context-aware handover decisions. A multi-tier simulation environment consisting of LTE, 5G NR, and WiFi 6 networks was developed using realistic path loss, shadowing, interference, and mobility models. Extensive evaluations demonstrated substantial improvements in mobility prediction accuracy, handover stability, radio link reliability, throughput efficiency, and latency reduction compared to conventional RSS-based and DQN-based schemes. The findings highlight the effectiveness of integrating predictive intelligence with reinforcement learning for reliable mobility management in 5G-Advanced and emerging 6G networks.

Big Data and Cognitive Computing doi: 10.3390/bdcc10050138

Authors: Latifa Boubekri Hassnae Aberkane Mohammed Chaouki Abounaima Loubna Lamrini

The TOPSIS (Technique for Order Preference by Similarity to Ideal Solution) method is one of the most widely used multi-criteria decision-making (MCDM) approaches in industrial, financial, and scientific fields. However, its sequential computational cost of O(m × n), where m denotes the number of alternatives and n the number of criteria, becomes prohibitive when decision matrices have several million rows. Despite its geometric interpretability and simplicity, classical TOPSIS faces two key computational bottlenecks at scale: (i) Euclidean distance calculations O(m × n) dominating the total cost, and (ii) the O(m × log m) sorting step, both inherently sequential and memory-bound on CPUs. To overcome these limitations, we propose GPU-TOPSIS, a fully vectorized and parallel reformulation of TOPSIS based on tensor execution on graphics processing units (GPUs), whose main contributions are: (i) a formally correct reformulation of TOPSIS as a GPU tensor pipeline preserving mathematical fidelity to the original method; (ii) a two-pass fragment-processing algorithm guaranteeing exact mathematical equivalence with monolithic TOPSIS, while reducing the memory footprint from O(m × n) to O(mt × n), where mt < m is the size of each independently processed fragment; (iii) three independent implementations on CuPy, PyTorch, and TensorFlow, ensuring the framework’s portability and genericity. Experimental evaluations on real data from the Amazon Products 2023 dataset, using matrices of up to 200 million alternatives (via the 2-pass formulation), demonstrate speedups of up to 4.75× compared to the reference CPU implementation (NumPy), with inter-backend score differences below 5 × 10−8 and 100% ranking overlap across all tested Top-K thresholds. A perturbation sensitivity analysis of the criteria weights and cross-backend consistency tests confirms that GPU acceleration fully preserves robustness and decision reliability, making GPU-TOPSIS a practical, open, and reproducible solution for large-scale multi-criteria decision making in Big Data environments.

Big Data and Cognitive Computing doi: 10.3390/bdcc10050137

Authors: Yuhan Zhang Zehang Wang Yuanfang Chen Chunfeng Du Jing Chen

With the continuous growth of data-intensive applications and artificial intelligence workloads, traditional dynamic random access memory (DRAM) is increasingly struggling to meet demands in terms of capacity scale, energy consumption constraints, and data retention after power failure. Consequently, non-volatile memory (NVM) has emerged as a crucial technology for bridging the gap between the memory and storage layers. However, due to inherent differences in write life, read–write performance variations, and consistency guarantee after failure, the systematic application of NVM still faces a series of challenges. Addressing these issues, this paper takes as its starting point the adaptation of medium characteristics and system design, and summarizes the research progress in aspects such as write optimization, consistency and security coordination mechanisms, data structure modification under hybrid memory architecture, and cross-layer resource collaboration. It also conducts an in-depth analysis of representative solutions and evaluation methods. The review results show that current research has shifted from improving a single performance bottleneck to multi-mechanism collaborative optimization. Various technical approaches have proven complementary in alleviating write amplification, enhancing persistence efficiency, and optimizing access patterns. This paper demonstrates that achieving stable and scalable application of NVM requires establishing a more systematic collaborative design concept between durability, security, and performance. As AI training workloads and big data analytics place increasing demands on memory bandwidth and persistence, the techniques surveyed here provide a foundational basis for next-generation memory-centric computing infrastructures.

Big Data and Cognitive Computing doi: 10.3390/bdcc10050136

Authors: Abdellah Rezoug Mohamed Bader-el-den

The proliferation of online fraud has resulted in substantial financial damage to individuals and organizations alike, with web phishing emerging as one of the most pervasive and harmful attack vectors. In response, this paper proposes the Stacking Ensemble Models Generator (SEMG), a URL-based phishing detection approach that leverages a multi-objective Genetic Algorithm to jointly optimize Precision and Recall in the selection and configuration of stacking ensemble models. An initial pool of base learners is trained on labeled datasets and subsequently evolved through genetic operators toward a globally optimal ensemble. Experimental evaluation across five datasets sourced from Mendeley and UCI repositories demonstrates that SEMG consistently surpasses individual base learners and compares favorably against existing methods, attaining 99.2% performance across all metrics on D2 while matching or exceeding state-of-the-art results on the remaining benchmarks. These outcomes underscore the framework’s robustness and its potential for deployment in real-world phishing detection systems.

Big Data and Cognitive Computing doi: 10.3390/bdcc10050135

Authors: Zilin Tian Xin Wang Yunfei Long Liguo Zhang

Adversarial transferability makes black-box attacks practical and exposes weaknesses of deep neural networks for computer vision, image recognition, and visual understanding. Among various transferability-enhancing methods, input transformation is one of the most effective strategies. However, existing methods often ignore the decoupling of style and semantics in the input image, as well as the need for customized transformation strategies, resulting in limited performance gains or suboptimal outcomes. In this paper, we propose a novel Fourier-based perspective for input transformation generalization in the context of vision adversarial attacks. The main observations are that the Fourier amplitude captures stylistic information and the phase encompasses richer semantics which are crucial for visual understanding. Motivated by this, we develop a Fourier-based strategy, which performs a stylistic transform and semantic mixup on the input examples to improve transferability. To avoid inconsistent semantics of augmented images for the surrogate model, we mix the original images with the augmentations to maintain semantic consistency and mitigate imprecise gradients. Extensive experiments on ImageNet-compatible datasets demonstrate that our method consistently outperforms existing input transformation attacks.

Big Data and Cognitive Computing doi: 10.3390/bdcc10050134

Authors: Alexander A. Karmanov Petr V. Nikitin

Energy-intensive refrigeration systems require predictive models that remain informative under counterfactual control trajectories, not only on archived operation. This paper develops a control-oriented multi-step state model and a nonlinear model predictive control framework for an indoor ice-rink refrigeration system. Historical state, control, and exogenous variables are encoded jointly with an admissible future control trajectory, and a normalized thermal-balance residual is added to the training objective. A lightweight conditioned transformer predicts ice temperature, return-glycol temperature, supply-glycol temperature, and compressor power over a 30 min horizon. The selected weakly regularized model with regularization coefficient λphys= 0.001 decreases the normalized thermal-balance root-mean-square error on the horizon tail by 30.29% relative to the base model while increasing the average ice-temperature root-mean-square error by only 1.90%. In a surrogate-based counterfactual four-day evaluation, the resulting nonlinear model predictive controller reduces predicted daily energy by 4.84%, terminal violation share by 17.32%, mean absolute terminal ice-temperature deviation by 18.74%, and the mean objective value by 30.82% relative to historical admissible setpoint tracking. The mean full control cycle time is 0.0311 s, confirming real-time feasibility for a 5 min supervisory update interval. All controller results are surrogate-based rather than field-deployed and therefore represent receding-horizon benchmark results under learned-model evaluation, not realized field savings.

Big Data and Cognitive Computing doi: 10.3390/bdcc10050133

Authors: Walaa Othman Batol Hamoud Nikolay Shilov Alexey Kashevnik Alexander Mayatin

Blood pressure monitoring is important for overall health assessment, yet traditional cuff-based methods are intrusive and unsuitable for continuous monitoring. This paper proposes a contactless approach for blood pressure estimation from facial videos using a bidirectional Mamba-based architecture with uncertainty quantification. Our method processes 64-frame video segments through a hierarchical 3D convolutional encoder to extract spatiotemporal features, then applies bidirectional state-space modeling to capture temporal dynamics efficiently. The model was evaluated on the Vitals for Vision (V4V) dataset, achieving mean absolute errors of 13.15 mmHg for systolic and 9.56 mmHg for diastolic blood pressure, outperforming prior methods while requiring significantly fewer computational resources than attention-based approaches. While these results do not meet clinical-grade diagnostic standards, they demonstrate the feasibility of contactless blood pressure estimation for non-clinical applications such as wellness monitoring, preliminary health screening, and continuous remote observation, where unobtrusive and computationally efficient monitoring is desirable.

Big Data and Cognitive Computing doi: 10.3390/bdcc10050132

Authors: Soufiyan Ouali Kanza Raisi Asmaa Mourhir El Habib Nfaoui Said El Garouani

Offensive language detection is crucial for ensuring safe and inclusive digital environments. Identifying harmful content protects users and supports healthier online interactions. Despite advances in transformer-based models, particularly Large Language Models (LLMs), their application to this task remains underexplored for low-resource languages such as Moroccan Arabic, especially compared with high-resource languages. This study evaluates the performance of various open- and closed-source LLMs for offensive language detection in Moroccan Darija. The evaluated models include general-purpose LLMs such as LLaMA, Mistral, and Gemma, as well as Arabic-focused models such as ArabianGPT, Falcon Arabic, and Atlas-Chat. We also experiment with reasoning models such as DeepSeek and GPT-4. Beyond traditional evaluation metrics, we investigate the robustness of these LLMs and examine the impact of adversarial training on their performance. Moreover, we contribute to the field by creating a large, high-quality dataset. Our evaluation revealed that GPT-4o Mini achieved the best overall performance, reaching an F1-score of 88%. However, robustness testing under black-box and white-box adversarial attacks exposed notable vulnerabilities, with attack success rates reaching 30%, thereby highlighting the need for enhancement. Despite the complex morphology and linguistic variability of Moroccan Darija, adversarial training resulted in a notable improvement in both overall model performance and robustness against adversarial attacks, yielding an average increase of 20.89% in resistance to attacks. Furthermore, this approach enabled GPT-4o Mini to achieve an F1-score of 91%, surpassing the current state-of-the-art performance by 6%. These results highlight the importance of incorporating adversarial approaches in low-resource dialectal settings to effectively address linguistic variability and data scarcity.

Big Data and Cognitive Computing doi: 10.3390/bdcc10050131

Authors: Areej Hamza Amel Tuama Asraf Mohamed Moubark

The high accuracy achieved by deep learning techniques has made them indispensable in computer vision applications. However, their substantial memory demands and high computational complexity limit their deployment in resource-constrained environments. To address this challenge, this study introduces a Feature Enhancement Module (FEM) as part of a unified hybrid compression framework that combines mixed-precision quantization and structured pruning to improve model efficiency. Experimental results on the Tiny ImageNet dataset using ResNet50 and MobileNetV3 architectures demonstrate the strong adaptability and scalability of the proposed approach. Compared with state-of-the-art compression methods, the proposed FEM-based framework achieves up to 6% improvement in Top-1 accuracy, while reducing memory usage by 32.26% and improving inference speed by 66%. Furthermore, the ablation study demonstrates that incorporating the FEM module leads to up to 24% improvement over the baseline model, highlighting its effectiveness. The results further show that FEM effectively preserves inter-channel feature representation stability even under aggressive compression, making it well suited for real-time processing and practical Artificial Intelligence (AI) applications. By maintaining semantic richness while significantly reducing computational cost, the proposed method bridges the gap between high-performance deep models and lightweight, deployable solutions. Overall, the FEM-based hybrid compression framework establishes a scalable and architecture-independent foundation for sustainable deep learning in resource-limited environments.

Big Data and Cognitive Computing doi: 10.3390/bdcc10050130

Authors: Zhiyuan Wang Tristan Lim Yun Teng Chongwu Xia

This study proposes a novel data-driven machine learning (ML) framework for multi-criteria environmental, social, and governance (ESG) evaluation. The framework aims to address the transparency, consistency, and subjectivity limitations of existing ESG evaluation systems by employing a fully data-driven, modular, and ML-supported architecture. It comprises three main modules: (i) ESG data preprocessing with missing-data imputation by the MissForest algorithm; (ii) a three-plane ESG feature selection workflow that integrates clustering, feature importance, and classification algorithms to identify representative ESG indicators; and (iii) a hybrid weighting and ranking procedure that combines unsupervised principal component analysis (PCA), criteria importance through inter-criteria correlation (CRITIC), and technique for order preference by similarity to ideal solution (TOPSIS) methods. A recent 2024 real-world application involving 57 listed Chinese pharmaceutical and biotechnology companies and 70 ESG indicators demonstrates the framework’s practical utility in producing transparent and objective ESG rankings. The main contributions of this work are fourfold: (1) the development of an end-to-end, entirely data-driven ML framework for ESG evaluation; (2) the introduction of an innovative three-plane ESG feature selection workflow within the framework; (3) the first study for designing a hybrid PCA-CRITIC-TOPSIS approach in ESG weighting and ranking; (4) the validation of the framework through a real-world industry application using recent and authentic ESG data.

Big Data and Cognitive Computing doi: 10.3390/bdcc10040129

Authors: Xiao Feng Shuaibing Lu Taotao Gu Yuanping Nie Qian Yan Mucheng Yang Jinyang Chen Xiaohui Kuang

Fuzzing is an essential element of software supply chain security governance. Despite its importance, the widespread adoption of library fuzzing is limited by the significant costs associated with constructing fuzz drivers. Without a clear entry point, the reachable path space of the target library is determined by the interplay of API call sequences, parameter dependencies, and state constraints. As a result, fuzz drivers must achieve not only successful builds but also provide sufficient semantic context to enable exploration of deeper state machine interactions, thereby avoiding premature stagnation at superficial validation logic. To systematically assess advancements in automated fuzz driver generation, this paper develops a taxonomy organized around the primary data sources used to derive driver-generation constraints, categorizing existing approaches into four technological trajectories: Usage Artifact Mining, Source Code Constraint Inference, Binary Semantics Recovery, and Heterogeneous Data Fusion. Large language models are increasingly integrated into these workflows as generators and as components for constraint alignment and repair. To address inconsistencies in experimental methodologies, this paper introduces a bounded comparability-oriented evaluation perspective focused on three dimensions: validity, reachability-related evidence, and reproducibility and cost. Together with a disclosure and reporting protocol for metric comparability, this perspective clarifies the information needed for cross-study comparison and examines the unique features and inherent limitations of each technical trajectory. Based on these findings, three key directions for future research are identified: facilitating structural evolution in response to coverage plateaus to address deep logic unreachability; coordinating dynamic closed-loop orchestration that utilizes on-demand heterogeneous data retrieval to resolve context challenges; and developing language-agnostic driver representations with pluggable adaptation mechanisms to improve cross-ecosystem portability and scalability.

Big Data and Cognitive Computing doi: 10.3390/bdcc10040128

Authors: Vismaya V. S Swetha P Jubin K. Babu Diya Gijo Varada M. T Adithya K. K Ekaterina Kopets Sishu Shankar Muni

Reservoir computing (RC) is an efficient framework for processing time-series data. This work investigates the synchronization of two independently trained reservoir computers that, after training, operate without external input from the chaotic system and interact solely through symmetric linear coupling. This approach addresses a gap in existing reservoir computing-based synchronization studies, which predominantly rely on master–slave or system-driven configurations. In this work, we first build and train two reservoir computing models based on 3D nonlinear chaotic maps and hyperchaotic systems and then introduce a symmetric linear coupling mechanism between them. This study demonstrates that reservoir computing can accurately reproduce the short-term dynamics of chaotic systems and provides insight into the interactions between learned dynamical models, while also helping us understand how complex systems connect and operate collectively. We use this systematic approach to establish a framework for understanding how two trained reservoir computers interact under varying coupling strengths, enabling a detailed investigation of their synchronization behavior. To demonstrate the adaptability of the proposed framework to diverse dynamical behaviors, we systematically investigated three discrete chaotic and hyperchaotic systems: (1) discrete 3D sinusoidal map with discrete Lorenz attractor, (2) 3D sinusoidal map with conjoined Lorenz twin attractor, and (3) 3D quadratic hyperchaotic map. For performance evaluation, we trained coupled RCs and computed the synchronization error for different coupling strengths. We also present phase portraits and time-series plots of the attractors and RCs, along with the synchronization error as a function of the coupling strength, thereby demonstrating the possibility of synchronization of two linearly coupled RCs, which are independently trained on discrete, three-dimensional chaotic and hyperchaotic systems.

Big Data and Cognitive Computing doi: 10.3390/bdcc10040127

Authors: Fabrizio Marozzo Riccardo Cantini

In recent years, generative artificial intelligence and, in particular, large language models (LLMs) have rapidly transformed the landscape of data analysis, knowledge extraction, content generation, and intelligent decision support [...]

Big Data and Cognitive Computing doi: 10.3390/bdcc10040126

Authors: Martin Moreno Iván Trejo-Zúñiga Víctor Alejandro González-Huitrón René Francisco Santana-Cruz Raúl García García Gabriela Pineda Chacón

Image-based AI offers a low-cost alternative to traditional turbidity sensors in farm ponds, yet the prevailing shift toward Vision Transformers (ViTs) critically overlooks two field realities: the chronic scarcity of annotated data (Small Data) and the strict computational limits of edge hardware. This study presents a frugal computer vision framework that challenges the need for complex architectures in environmental screening. By systematically benchmarking six deep learning models across a calibrated high-turbidity dataset (200–800 NTU, 700 images) under standardized capture conditions, we demonstrate that traditional Convolutional Neural Networks (CNNs) possess a crucial inductive bias for this task. Specifically, ResNet-50 significantly outperformed modern ViTs in both accuracy (96.3% vs. 80.0%) and data efficiency, effectively capturing spatial scattering patterns without the massive data requirements that hindered transformer convergence. Deployed on a resource-constrained Raspberry Pi 4, the CNN-based system achieved an inference latency of 46 ms, demonstrated in an initial hardware-in-the-loop field proof-of-concept (82.4% agreement under baseline, calm-weather conditions, n=17). This edge-native approach not only provides actionable spatial turbidity maps to guide on-farm filtration and livestock management decisions but also establishes a critical architectural baseline: under controlled capture protocols, mature CNNs consistently outperform ViTs, establishing them as the optimal architecture for frugal, small-data agricultural Internet of Things (IoT) deployments.

Big Data and Cognitive Computing doi: 10.3390/bdcc10040125

Authors: Khadija Lasri Khalid El Fazazy Adnane Mohamed Mahraz Hamid Tairi Jamal Riffi

While Graph Convolutional Networks (GCNs) have revolutionized skeleton-based action recognition, existing methods face a critical efficiency–accuracy dilemma: state-of-the-art approaches achieve high performance through computationally expensive multi-stream fusion (joint, bone, joint motion, and bone motion) and deep architectures, limiting real-world deployment on resource-constrained devices. We propose LST-AGCN (Lightweight Spatial–Temporal Attention Graph Convolutional Network), introducing three technical contributions that address this challenge: (1) Unified Attention Module (UAM)—a framework that integrates channel, spatial, and temporal attention through a single compact operation, significantly reducing attention parameters compared to separate attention mechanisms; (2) Depthwise Separable Attention Mechanism (DSAM)—a factorization using depthwise separable convolutions that achieves linear complexity reduction from O(C2) to O(C) in attention operations; and (3) Efficient Topology-Aware Fusion (ETAF)—an adaptive Joint-wise Attention strategy that captures fine-grained spatial relationships without quadratic complexity growth. Extensive experiments on NTU RGB+D 60 and NTU RGB+D 120 datasets demonstrate that LST-AGCN achieves strong performance using only joint modality (86.14%/94.0% and 79.5%/82.0% Top-1 accuracy with 99.0% Top-5 on cross-view) while requiring 14.11 M parameters and 19.02 GFLOPs, delivering efficient inference suitable for edge deployment.

Big Data and Cognitive Computing doi: 10.3390/bdcc10040124

Authors: Mabel López Bordao Antonia Ferrer Sapena Carlos A. Reyes Pérez Enrique A. Sánchez Pérez

This study aims to identify and synthesize the major global trends that shaped 2025 by applying the DeflyCompass methodology to a curated corpus of strategic foresight reports. The study synthesizes insights from 23 strategic reports published by leading international organizations, including the World Economic Forum, Accenture, Euromonitor, and major technology firms. Methodologically, DeflyCompass operationalizes a structured hybrid human–AI pipeline comprising the deployment of multi-agent AI systems, automated knowledge graph construction, semantic clustering, and hybrid human–AI validation processes, reducing an initial set of 816 preliminary signals to a validated catalog of 50 high-priority trends across six PESTEL domains: Political, Economic, Social, Technological, Environmental, and Legal/Governance. Key findings indicate that artificial intelligence functions as a systemic enabling technology across all domains, climate and sustainability imperatives permeate multiple domains, geopolitical fragmentation introduces systemic tension, and trust deficits emerge as a critical vulnerability. The study contributes a replicable and scalable framework for global-level strategic foresight that operationalizes human–AI integration within a rigorous expert-driven validation process, complementing existing hybrid analytical approaches in the literature. Implications extend to decision-making in technology governance, sustainability strategy, social adaptation, and scenario planning, highlighting the necessity of integrating AI augmentation with human expertise for effective future-oriented planning.

Big Data and Cognitive Computing doi: 10.3390/bdcc10040123

Authors: Ahsan Rafiq Eduard Melnik Alexey Samoylov Alexander Kozlovskiy Irina Safronenkova

As industries deploy more connected devices in factories, warehouses, and smart facilities, the need for artificial intelligence (AI) systems that can operate securely in distributed, data-intensive environments is growing. Traditional centralized learning and online education platforms struggle when students and systems have to process real-time streams (sensors, video, text) with strict latency and privacy requirements. To address this challenge, a blockchain-secured, edge-enabled multimodal federated learning framework tailored for Industrial IoT (IIoT) environments is proposed. The model integrates four key layers: (i) a blockchain layer that provides credentialing, transparency, and token-based incentives; (ii) a multimodal community layer that supports group formation, peer consensus, and cross-modal learning across text, images, audio, and sensor data; (iii) an edge computing layer that enables low-latency task offloading and secure training within Intel SGX enclaves; and (iv) a data layer that applies pre-processing, differential privacy, and synthetic augmentation to safeguard sensitive information. Experiments on industrial multimodal datasets demonstrate 42% faster model aggregation, 78.9% multimodal accuracy, and 1.9% accuracy loss under ε = 1.0 differential privacy. This shows a scalable and practical path for decentralized AI training in next-generation IIoT systems, confirming the possibility of technical support for educational processes. However, the conducted research requires a validation of pedagogical effectiveness.

Big Data and Cognitive Computing doi: 10.3390/bdcc10040122

Authors: Mikhail Rumiantcev

Analyzing music similarity in large catalogs is challenging because people perceive music differently and important details are found in audio, text, and metadata. This article introduces a multimodal framework that uses an ontology to make music similarity and recommendation more explainable. The framework brings together learned features from audio, lyrics, and other text with structured metadata in a shared similarity space, and then improves ranking with a music ontology that captures relationships between songs, artists, genres, and moods. The design works with any encoder that creates fixed-size features. This study uses strong neural audio and text encoders, mainly based on transformers. This approach allows the system to handle different input types while staying reliable across datasets. This study tests the framework on several open music and audio datasets using content-based retrieval tasks and standard ranking measures. In addition to Configurations C1–C4, this study includes an external content-based reference baseline based on conventional MIR audio descriptors. This baseline represents a signal-level retrieval approach that models complementary aspects of the audio signal, such as timbre, harmony, and spectral characteristics, and is evaluated under the same retrieval protocol as the main framework. It is included to provide an external comparison point outside the proposed C1–C4 design. Compared to audio-only and non-ontological variants within the same framework, the proposed multimodal and ontology-guided configurations achieve better precision, recall, and mean average precision, and also cover more rare content. Visualizations and case studies show that combining different data types and using ontology-based reranking can improve performance and make results easier to interpret. This work lays the groundwork for explainable, cognitively informed music recommendation systems and points to future work in modeling user behavior over time and adapting to different cultures.

Big Data and Cognitive Computing doi: 10.3390/bdcc10040121

Authors: Noé Y. Flandre Philippe J. Giabbanelli

Communicating the design and results of agent-based models (ABMs) to subject matter experts is challenging, which hinders participation and limits trust in simulation-based decision support. Large language models (LLMs) can communicate ABMs as textual summaries, thus complementing traditional disclosure through statistical and visualization techniques. While prior work translated the structure of conceptual models into narratives via LLMs, our extension covers the dynamics of simulation models via an automated simulation-to-text method that extracts contextual information from NetLogo ABMs, performs repeated simulations, and generates narrative descriptions (including the model’s purpose, parameters, and simulation dynamics) using mutimodal LLMs. Furthermore, four summarization algorithms spanning abstractive and extractive methods provide shorter reports. Using Design-of-Experiments methods over three peer-reviewed ABMs, state-of-the-art multimodal LLMs from 2026 (Gemini 3.1 Pro, Qwen 3.5, Kimi K2.5, Claude Opus 4.6) and different prompt elements (e.g., roles, examples, generating insights, statistical analyses), we compare our results with several reference reports (e.g., from associate professors). We find that report quality is determined mainly (i.e., up to 34% of the variance) by the summarization algorithm and its interaction with the LLM, with abstractive summarizers (BART, T5) producing more coherent and readable reports, while Claude Opus 4.6 is the most robust LLM.

Big Data and Cognitive Computing doi: 10.3390/bdcc10040120

Authors: Francesco Granata Francesco Poggi Misael Mongiovì

In the era of Large Language Models (LLMs), Retrieval-Augmented Generation (RAG) architectures are gaining significant attention for their ability to ground language generation in reliable knowledge sources. Despite their effectiveness, RAG systems based solely on semantic similarity often fail to ensure factual accuracy in specialized domains, where terminological ambiguity can affect retrieval relevance. This study proposes Entity Linking Enhanced RAG (ELERAG), an enhanced RAG architecture that integrates a factual signal derived from Entity Linking to improve the accuracy of educational question-answering systems in Italian. The system includes a Wikidata-based Entity Linking module and implements a hybrid re-ranking strategy based on Reciprocal Rank Fusion (RRF). To validate our approach, we compared it against standard baselines and state-of-the-art methods, including a Weighted-Score Re-ranking, a standalone Cross-Encoder and a combined RRF + Cross-Encoder pipeline. Experiments were conducted on two benchmarks: a custom academic dataset and the standard SQuAD-it dataset. Results show that, in domain-specific contexts, ELERAG significantly outperforms both the baseline and the Cross-Encoder configurations. Conversely, the Cross-Encoder approaches achieve the best results on the general-domain dataset. These findings provide strong experimental evidence of the domain mismatch effect, highlighting the importance of domain-adapted hybrid strategies to enhance factual precision in educational RAG systems without relying on computationally expensive models trained on disparate data distributions. They also demonstrate the potential of entity-aware RAG systems in educational environments, fostering adaptive and reliable AI-based tutoring tools.

Big Data and Cognitive Computing doi: 10.3390/bdcc10040118

Authors: Kosmas Katsioulas Ilias Maglogiannis

Handball performance analysis is still often conducted through the manual review of match videos, while automation on broadcast footage remains challenging due to camera motion, strong perspective effects, and frequent occlusions during dense interactions. This study presents a practical and reproducible monocular pipeline for extracting handball analytics from a single broadcast viewpoint. Players are detected per frame, tracked over time, and projected onto a standardized handball court via homography-based camera calibration. The resulting court-referenced trajectories in metric units enable motion indicators such as distance covered and speed, along with coaching-oriented visual summaries, including trajectory overlays and heatmaps. In addition, clip-level action recognition is performed using interpretable kinematic and scene-derived features and lightweight classifiers, with a comparative evaluation across multiple classical models. The modular design keeps the intermediate steps explicit, supports reproducibility, and facilitates interpretation of both intermediate outputs and final analytics. Experiments on the UNIRI handball dataset demonstrate that meaningful performance analytics and action understanding can be obtained from single-camera broadcast video using transparent intermediate representations. This work highlights the practical potential of interpretable trajectory-based modeling for under-instrumented sports and provides a reproducible baseline for future extensions incorporating richer contextual cues.

Big Data and Cognitive Computing doi: 10.3390/bdcc10040119

Authors: Naima Firdaus Sachin Balkrushna Jadhav Zahid Raza Maria Lapina Mikhail Babenko

Federated learning has become a promising paradigm in the training of decentralized machine learning models across clients without sharing raw data, thereby preserving privacy. Current federated support vector machine methods are mainly based on the learning of a single global model, which inadequately addresses the challenges presented by heterogeneous and non-IID client data distributions. To overcome these limitations, we propose FMT-SVM, a novel federated multi-task learning framework that jointly trains both binary and multi-class classification tasks within each client, where the client uses a unified convolutional neural network encoder to extract common features among tasks, which are passed to task-specific linear SVM heads dedicated to each classification task. By leveraging a primal optimization integrating task covariance and global consensus regularization, FMT-SVM explicitly models relationships between heterogeneous tasks and enforces alignment across clients, effectively handling the non-IID nature of data distributions. Unlike traditional FL methods that exchange entire model parameters or large support vector sets, our method communicates only the compact SVM heads during aggregation, greatly reducing communication overhead and enhancing scalability for clients with limited bandwidth. To further enhance privacy, Gaussian differential privacy mechanisms are applied to client updates, balancing privacy preservation with predictive performance. Experiments are performed on two medical image datasets: the Pediatric Pneumonia Dataset and the Breast Ultrasound dataset, demonstrating that the FMT-SVM framework achieves competitive accuracy on both binary and multi-class tasks while maintaining communication efficiency and privacy guarantees. These results highlight the capability of the proposed FMT-SVM framework as a practical, scalable, and privacy-aware solution for the federated true multi-task learning problem in sensitive healthcare applications.

Big Data and Cognitive Computing doi: 10.3390/bdcc10040117

Authors: Anna Dalla Vecchia Simone Mattioli Sara Migliorini Elisa Quintarelli

Understanding and predicting how tourists move through a city is a challenging task, as it involves a complex interplay of spatial, temporal, and social factors. Traditional recommender systems often rely on structured data, trying to capture the nature of the problem. However, recent advances in Large Language Models (LLMs) open new possibilities for reasoning over richer, text-based representations of user context, even without a dedicated pre-training phase. In this study, we investigate the potential of LLMs to interpret and predict tourist movements in a real-world application scenario involving tourist visits to Verona, a municipality in Northern Italy, between 2014 and 2023. We propose an incremental prompt engineering approach that gradually enriches the model input, from spatial features alone to richer behavioral information, including visit histories, time information, and user cluster patterns. The approach is evaluated using six open-source models, enabling us to compare their accuracy and efficiency across various levels of contextual enrichment. The results provide a first insight about the abilities of LLMs to incorporate spatio-temporal contextual factors, thus improving predictions, while maintaining computational efficiency. The analysis of the model-generated explanations completes the picture by adding an interpretability dimension that most existing next-PoI prediction solutions lack. Overall, the study demonstrates the potential of LLMs to integrate multiple contextual dimensions in tourism mobility, highlighting the possibility of a more text-oriented, adaptive, and explainable T-RS.

Big Data and Cognitive Computing doi: 10.3390/bdcc10040115

Authors: Vasiliy Pchelko Vladislav Kholkin Vyacheslav Rybin Alexander Mikhailov Timur Karimov

Despite the success of artificial neural networks in solving numerous tasks, they face significant challenges, including difficulties in online adaptation and rapidly increasing energy consumption. As a biologically plausible alternative, spiking neural networks offer promising capabilities for efficient cognitive computing. Recently, a three-element spiking neuron model consisting of a threshold selector, a tunnel diode, and a capacitor was proposed. In this work, we experimentally validate this model using a threshold selector hardware emulator and demonstrate its dynamical equivalence to the biologically plausible Izhikevich neuron model. To evaluate the novel neuron’s applicability for cognitive computing, we implement a liquid state machine (LSM) reservoir architecture with spatially dependent random topology for synaptic weight distribution. Our simulations on the MNIST and Fashion-MNIST benchmarks demonstrate competitive classification accuracy (97.9% and 89.5%, respectively) while offering estimated energy efficiency and processing speed enhancements compared to existing FPGA-based and memristor-based spiking reservoir implementations. The developed reservoir is feasible for processing neuromorphic sensors output, including visual perception tasks.

Big Data and Cognitive Computing doi: 10.3390/bdcc10040116

Authors: Luisiana Sabbatini Alberto Belli Sara Bruschi Marco Esposito Sara Raggiunto Paola Pierleoni

Sleep plays a fundamental role in human health and cognitive functioning, motivating the development of reliable and scalable methodologies for sleep stage classification (SSC). Recent advances in non-invasive and economically sustainable sensing technologies enable continuous sleep monitoring beyond laboratory settings. However, SSC remains a challenging data analytics task due to the intrinsic class imbalance among sleep stages. This study investigates the effectiveness of different imbalanced data management strategies within a machine learning framework for non-invasive SSC. The proposed approach relies exclusively on heart rate and motion signals, which can be acquired through wearable devices or contactless under-mattress sensors, making it suitable for longitudinal monitoring scenarios. Using the PhysioNet DREAMT dataset, 32 experimental scenarios are defined by combining data-level techniques (ADASYN oversampling with different balancing weights), algorithm-level strategies (cost-sensitive learning), and hybrid solutions. Four model families are evaluated—Decision Tree, k-Nearest Neighbors, Ensemble Classifiers, and Artificial Neural Networks—across classification tasks involving 2, 3, 4, and 5 sleep stages. The experimental results show that ensemble-based models provide robust and consistent performance under severe class imbalance, achieving macro accuracies of 82% for sleep–wake detection, 73% for 3-stage classification, 72% for 4-stage classification, and 64% for 5-stage classification. These findings confirm the relevance of imbalance-aware analytics and demonstrate the feasibility of accurate, minimally invasive SSC within big data and cognitive computing paradigms.

Big Data and Cognitive Computing doi: 10.3390/bdcc10040114

Authors: Irina Evgenievna Kalabikhina Anton Vasilyevich Kolotusha Vadim Sergeevich Moshkin

Patients leave millions of medical reviews annually, providing critical data for quality management. However, manual processing is infeasible, and existing systems fail to distinguish medical from organizational problems—a distinction essential for complaint routing. The consequences of misrouting are significant: clinical issues may go unaddressed when medical complaints reach administrative staff, while systemic service problems remain unresolved when organizational complaints reach medical directors. We developed a hybrid approach combining expert annotation with Large Language Models (LLMs). Fifteen prompt iterations on 1500 reviews with expert validation (modified Cohen’s kappa (κ_mod), which weights errors hierarchically, reached 0.745) preceded the LLM annotation of 15,000 mixed-sentiment and positive reviews. These were combined with 7417 expert-annotated negative reviews to form a corpus of 22,417 reviews. Eight architectures, ranging from Logistic Regression to a BERT + TF-IDF + LightGBM ensemble, were compared using both standard metrics and domain-specific practical metrics tailored to complaint routing. The best model, scaled to 4.3 million Russian-language reviews from the Prodoctorov.ru platform, achieved 92.9% Practical Accuracy—the proportion of reviews classified without critical medical–organizational misclassification errors (M ↔ O)—compared to 68.0% standard accuracy, which treats all errors equally. Critical errors were reduced to 1.4%, yielding 144,000 more correctly processed complaints than traditional methods (TF-IDF + Logistic Regression). Analysis of the scaled data revealed the following: 46.1% M (medical), 21.0% O (organizational), and 32.9% C (combined) reviews; medical ratings were highest (4.75 vs. 4.59 for organizational, p < 0.001); combined reviews were longest (802 characters); zero-star reviews comprised 3.8% of feedback, with organizational complaints dominating (38.2%) among extreme negatives; and average ratings rose by 1.24 points over 14 years. This hybrid approach yields expert-comparable corpora, automates 93% of feedback processing, ensures correct complaint routing, and contributes to healthcare sustainability by reducing administrative burden, accelerating resolution, and enabling data-driven quality management without proportional increases in human resources. All analyses were conducted on Russian-language patient reviews.

Big Data and Cognitive Computing doi: 10.3390/bdcc10040113

Authors: Diego Labate Dipanwita Thakur Giancarlo Fortino

Smart-meter theft detection requires learning from fine-grained electricity consumption data, whose centralized processing poses significant privacy risks. Federated learning (FL) mitigates these risks by decentralizing training, but providing rigorous user-level differential privacy (DP) under non-IID data and heterogeneous client behavior remains challenging. Existing DP-FL approaches rely on fixed global clipping bounds for client updates, which substantially overestimate sensitivity when privacy loss is composed using Rényi Differential Privacy (RDP), zero-Concentrated DP (zCDP), or Moments Accountant (MA) frameworks, leading to excessive noise and degraded utility. This work proposes an adaptive clipping-based RDP accountant that incorporates empirical, round-wise update magnitudes into privacy accounting by rescaling each round’s RDP contribution according to the observed clipping ratio. The method is optimizer-agnostic and is evaluated with FedAvg, FedProx, and SCAFFOLD on the SGCC smart-meter theft dataset under IID and Dirichlet non-IID partitions. Experimental results show consistently tighter privacy bounds and improved model utility compared to classical DP accountants, demonstrating the effectiveness of sensitivity-aware privacy accounting for practical differentially private FL.

Big Data and Cognitive Computing doi: 10.3390/bdcc10040112

Authors: Paulino José García-Nieto Esperanza García-Gonzalo José Pablo Paredes-Sánchez Luis Alfonso Menéndez-García

The higher heating value (HHV), sometimes referred to as the gross calorific value, is a crucial metric for determining a fuel’s primary energy potential in energy production systems. By combining extreme gradient boosting (XGBoost) with the differential evolution (DE) optimizer, an innovative machine learning-based model was created in this study to forecast the HHV (dependent variable). As input variables, the model included the constituents of the coal’s ultimate analysis: carbon (C), oxygen (O), hydrogen (H), nitrogen (N), and sulfur (S). For comparative purposes, random forest regression (RFR), M5 model tree, multivariate linear regression (MLR), and previously reported empirical correlations were also applied to the experimental dataset. The results showed that the XGBoost strategy produced the most accurate predictions. An initial XGBoost analysis was carried out to identify the relative contribution of the input variables to coal HHV prediction. In particular, for coal HHV estimates reliant on experimental samples, the XGBoost regression produced a correlation coefficient of 0.9858 and a coefficient of determination of 0.9691. The excellent agreement between observed and anticipated values shows that the DE/XGBoost-based approximation performed satisfactorily. Lastly, a synopsis of the investigation’s key conclusions is provided.

Big Data and Cognitive Computing doi: 10.3390/bdcc10040111

Authors: Thaer AL Ibaisi Stefan Kuhn Muhammad Kazim Ismail Kara Turgay Altindag Mujeeb Ur Rehman

The increasing deployment of Internet of Things (IoT) devices introduces significant security challenges, while privacy concerns limit centralized data aggregation for intrusion detection. Federated learning (FL) offers a decentralized alternative, yet the interaction between feature representation, model architecture, and data heterogeneity remains insufficiently understood in IoT malware detection. This study provides a controlled comparative analysis of centralized and federated learning, optionally using amino acid encoding, under IID and Non-IID conditions using a 10,000-sample subset of the CTU–IoT–Malware–Capture dataset. First, we evaluate raw tabular features versus amino acid-based feature encoding, followed by a lightweight multi-layer perceptron (2882 parameters) versus a deeper residual network (70,532 parameters), across binary and multi-class classification tasks. In the binary setting, centralized training achieved up to 98.6% accuracy, while federated IID training reached 98.6%, with differences within statistical variance. Under Non-IID conditions, performance decreased modestly (0.1–0.5 percentage points), and accuracy was consistently lower when using encoded features compared with raw features. The degradation is smaller in deeper architectures and may offer improved stability under highly skewed federated conditions. In the four-class setting, the complex network achieved up to 97.8% accuracy with raw features, while amino acid encoding achieves up to 93.3%. The results show that federated learning can achieve performance comparable to centralized training under moderate heterogeneity, that lightweight architectures are sufficient for low-dimensional IoT traffic features, and that feature compression via amino acid encoding does not inherently mitigate Non-IID effects. These findings clarify the relative impact of representation, heterogeneity, and architectural capacity in practical FL-based IoT intrusion detection systems.

Big Data and Cognitive Computing doi: 10.3390/bdcc10040110

Authors: Leonidas Theodorakopoulos Aristeidis Karras Alexandra Theodoropoulou Christos Klavdianos

Enterprise decision making in marketing, finance, and audit remains fragmented, leading to inefficient budget allocation and incomplete risk assessment. This study proposes an integrated, Big Data-driven decision-support framework that unifies Large Language Models (LLMs), attention-based marketing mix modeling, and multi-agent, game-theoretic optimization to coordinate cross-functional decisions. The architecture combines five modules: LLM-enhanced customer segmentation and customer lifetime value prediction, attention-weighted marketing mix modeling, multi-agent LLM systems for hierarchical budget optimization, attention-informed Markov multi-touch attribution, and LLM-augmented audit quality assessment. Empirical validation on a large-scale e-commerce dataset with 2.8 million customers and USD 156 million in marketing expenditure shows that marketing return on investment increases from 4.2 to 6.78 (61.4% relative improvement), financial forecasting error (MAPE) decreases from 12.8% to 4.7% (63.3% reduction), fraud detection accuracy improves by 29.8%, the Audit Quality Index reaches 0.951, and customer lifetime value prediction accuracy improves from 76.4% to 91.3%. By operationalizing the convergence of LLMs, attention mechanisms, and game-theoretic reasoning within a unified and empirically validated framework, the study delivers both theoretical advances and practically deployable tools for integrated business intelligence in digital economies.

Big Data and Cognitive Computing doi: 10.3390/bdcc10040109

Authors: Yu-Kai Huang Chih-Hung Chen Yun-Cheng Tsai Shun-Shii Lin

This study proposes a machine learning framework tailored to the institutional characteristics of Taiwan’s stock market, aiming to enhance forecasting accuracy for the Taiwan Stock Exchange Capitalization Weighted Stock Index (TAIEX). The model employs the Radius Neighbors Regressor with a dynamic radius-based similarity measure and integrates domain-specific features including technical indicators, volume–price relationships, and Qualified Foreign Institutional Investor (QFII) activity. A custom 60-day input window with a 20-day forecast horizon is applied to capture medium-term market dynamics. The framework was evaluated through extensive backtesting and real-world validation with the TAIEX Futures. The results demonstrate that the model achieves a peak directional accuracy of 85.1% under optimal parameter settings. Moreover, trading simulations confirm its practical viability, yielding a cumulative return on investment (ROI) of approximately 1600% during the short-term evaluation period (2023–2025) and nearly 2000% in the long-term evaluation (2019–2025), even after accounting for transaction costs and stop-loss mechanisms. These findings indicate that combining historical pattern similarity with institutional investor behavior substantially improves predictive power and profitability. Nevertheless, the framework remains constrained by its reliance on Taiwan-specific institutional features and historical trading data, limiting generalizability. Future research should extend applications to other markets while incorporating macroeconomic variables, corporate fundamentals, and news-driven signals to enhance adaptability.

Big Data and Cognitive Computing doi: 10.3390/bdcc10040108

Authors: Qing Chen Yujie Xue Jie Lin Chang Zhu

The rapid development of Generative Artificial Intelligence (GenAI) in graduate education has changed human–AI interaction within knowledge-intensive environments, leading to important questions about user-side cognitive adaptation in probabilistic AI systems. While many studies focus on ethical implications, limited attention has been paid to the cognitive mechanisms underlying graduate students’ adoption of GenAI. Drawing on the Technology Acceptance Model (TAM), this study explores the cognitive and interactional mechanisms shaping graduate students’ adoption and usage of GenAI. Using thematic analysis of in-depth interviews with 20 graduate students from diverse academic backgrounds, the study identifies seven interrelated constructs: perceived usefulness, perceived ease of use, external environment, risk perception, attitude, behavioral intention, and interaction subjectivity. This study demonstrates that the adoption of GenAI is not merely a result of perceived efficiency but is shaped by cognitive calibration between trust and risk evaluation. Moreover, interaction subjectivity emerges as a metacognitive factor that determines whether engagement results in human–AI collaboration or passive automation. By integrating external environment, risk perception, and interaction subjectivity, this study provides a cognitively grounded framework for understanding human–AI adoption and interaction dynamics. Practically, the findings provide design-relevant insights for developing GenAI systems that support calibrated trust, uncertainty awareness, and adaptive cognitive participation.

Big Data and Cognitive Computing doi: 10.3390/bdcc10040107

Authors: Jinsong Zhang Xiongzheng Li Yuqin Lin

Exemplar-based image translation generates an output image by transferring appearance from a reference exemplar to a content image. Existing works only consider the local correspondences between two modalities, and ignore the global distributions in each modality, struggling to obtain fine-grained details with efficient computation. In this paper, we propose OTFormer, a multi-scale Optimal Transport transformer for exemplarbased image translation. We formulate cross-modal alignment as a multi-scale optimal transport problem, which progressively provides a globally coherent matching. In addition, we design a lightweight multi-scale fusion block to extract and fuse features efficiently. Experiments on CelebA-HQ and DeepFashion demonstrate that OTFormer improves both image fidelity and style adherence, while reducing model parameters by 62% and achieving faster inference compared with strong baselines. These results highlight OTguided global alignment as an effective and deployable solution for high-fidelity exemplarbased image translation.

Big Data and Cognitive Computing doi: 10.3390/bdcc10040106

Authors: Kenan Kassab Alexey Kashevnik Irina Shoshina

Candidate interview assessment is primarily reliant on subjective human judgment, while existing AI-based methods rely on end-to-end predictions with no psychometric basis. In this paper, we propose an interpretable multi-modal framework that combines nonverbal behavior, LLM-based verbal analysis, and Big Five personality traits into three theory-based constructs: professional-cognitive competence, observed leadership behavior, and leadership disposition. The proposed method utilizes computer vision and larger language models to extract features from video interviews. Rather than targeting predictive accuracy, the proposed method prioritizes construct validity and transparent aggregation under severe label scarcity. The proposed method aggregates the constructs into a Top Potential Score that reflects the executive abilities of the candidate. Experiments on the method show its ability to significantly differentiate top candidates from others (Cliff’s delta = 0.91 for the composite Top Potential Score, permutation p = 0.0002). Leave-one-out analysis verifies robustness, while rank-based evaluation yields 100% recall of executive candidates in the top 20% of rated applications. The findings justify the use of the proposed multi-modal method as an interpretable decision-support tool for candidate interview assessment.

Big Data and Cognitive Computing doi: 10.3390/bdcc10040105

Authors: Jinsong Zhang Cheng Guan Zhihua Lin Yuqin Lin

Reconstructing photorealistic and animatable whole-body avatars from monocular videos is a hot topic in computer vision and computer graphics. However, existing methods still face challenges due to the limited frequency response of single-scale geometry encodings and the instability of appearance modeling without an explicit surface anchor. In this paper, we present H2Avatar, a real-time framework that builds on a mesh-embedded 3D Gaussian representation guided by SMPL-X and disentangles geometry and appearance into hierarchical and hybrid components. For geometry, we propose a semantic-aware hierarchical encoding based on a multi-scale tri-plane pyramid, where features at different resolutions capture both global structure and high-frequency surface details such as clothing wrinkles. For appearance, we introduce a hybrid rendering strategy that anchors canonical colors using a learnable UV texture map, and complements it with a neural residual color branch conditioned on tri-plane features, pose embedding, and surface normals to model pose- and view-dependent shading variations. This design improves temporal stability and preserves identity details while enhancing photorealism under complex motions. Experiments on the NeuMan dataset demonstrate that H2Avatar consistently outperforms representative baselines across multiple sequences, outperforming ExAvatar by up to 0.66 dB in PSNR and reducing LPIPS by up to 16.3%. These results validate the effectiveness of hierarchical geometry encoding and texture-anchored hybrid appearance modeling.

Big Data and Cognitive Computing doi: 10.3390/bdcc10040104

Authors: Linfei Ma Dexing Wei Xiangpeng Li Feng Wen Haisu Zhang

Schema linking, the task of identifying relevant database schema elements (tables and columns) for natural language queries, is a critical component in database-driven natural language interfaces. While existing approaches rely on question decomposition to handle complex queries, they often suffer from error propagation and low precision. In this paper, we propose a novel schema linking framework enhanced by self-verification (SV) and value hints (VHs) that significantly improves both precision and recall. Our approach introduces two key components: (1) self-verification (SV), an iterative refinement mechanism that validates and corrects initial predictions through explicit verification prompts, and (2) value hints (VHs), which explicitly guide the model to recognize database values mentioned in queries. We conduct comprehensive experiments on two benchmark datasets, Spider and BIRD, using two language models of 4B and 80B parameters. Our results demonstrate that SV + VH consistently improves performance across datasets, models, and method configurations, outperforming both decomposition-based approaches and compute-matched alternatives such as self-consistency under equivalent inference budgets.

Big Data and Cognitive Computing doi: 10.3390/bdcc10040103

Authors: Joel Lehmann Tim Markus Häußermann Julian Reichwald

Complex products and production processes are intertwined and demand expressive, lifecycle-wide digital representations. The Asset Administration Shell emerged as a standard for Digital Twins (DTs), structuring heterogeneous data across cloud-based Industrial Internet of Things (IIoT) infrastructures. However, today’s deployments predominantly realize passive or reactive DTs, while intelligent behavior remains underexploited. This paper addresses this gap, proposing an end-to-end architecture operationalizing the DT Reference Model through the integration of machine-interpretable granulated industrial skills, which are semantically accumulated into a knowledge graph enabling discovery and reasoning, while a multi-agent system provides autonomous, utility-based negotiation via machine-to-machine interactions within a federated marketplace. The approach is applied in a real smart manufacturing demonstrator, combining order processes, production orchestration, and lifecycle documentation into a unified execution pipeline spanning IIoT-connected shopfloor assets and cloud-based services. Quantitative experiments evaluating negotiation latency, renegotiation robustness, and utility variation demonstrate stable, predictable behavior even under concurrent demand and failure scenarios. The architecture lays a foundation for interoperable, sovereign collaboration across value chains to realize shared production. The results underline the effectiveness of the tightly coupled enabler technologies realizing proactive, reconfigurable, and semantically enriched intelligent DTs.

Big Data and Cognitive Computing doi: 10.3390/bdcc10040102

Authors: Manuel Gozzi Francesca Fallucchi

Large Language Models (LLMs) demonstrate remarkable performance across a variety of natural language understanding tasks, yet their sensitivity to emotional framing in user prompts remains underexplored. This paper presents an empirical study investigating how four emotional tones—joy, apathy, anger, and fear—affect LLM performance on the SuperGLUE benchmark. We evaluate five instruction-tuned, open-weight models across eight diverse tasks, systematically modulating input prompts with affective cues while keeping semantic content constant. Results reveal that prompts framed with joy and apathy lead to consistently higher accuracy, with gains of up to 4.5 percentage points compared to fear-framed inputs, which yield the lowest performance. These findings demonstrate that affective modulation in user prompts measurably impacts LLM reasoning and task outcomes, suggesting that emotional framing is not merely stylistic but functionally relevant to model behavior. Our study provides a reproducible experimental framework and an open-source prompt set, offering a foundation for future research on affect-aware prompting strategies and their implications in human–AI interaction.

Big Data and Cognitive Computing doi: 10.3390/bdcc10040101

Authors: Ding-Wei Chen Yun-Nan Chang

Atopic dermatitis (AD) is a chronic inflammatory skin disorder that is significantly contributed to by epigenetics. We developed a machine learning-based framework to identify DNA methylation biomarkers associated with AD classification and severity. Genome-wide methylation data from peripheral blood were processed using four feature selection algorithms: coarse approximation linear function (CALF), elastic net (EN), minimum redundancy maximum relevance (mRMR), and recursive feature elimination with cross-validation (RFECV). The integrative framework identified a central panel of 8 CpG sites that achieved an area under the curve (AUC) of 1.00 in the test set. This panel demonstrated high disease specificity, showing poor classification performance for systemic lupus erythematosus (AUC = 0.46), Crohn’s disease (AUC = 0.50), and oral squamous cell carcinoma (AUC = 0.58). Severity prediction using RFECV-selected 63 CpG sites (RFE63) achieved high accuracy across classifiers, with Random Forest (accuracy = 0.94) outperforming the others. The functional enrichment of CpG-associated genes highlighted key immune-related transcriptional regulators, including STAT5A, RUNX1, MEIS1, and PAX4. These genes are linked to chromatin remodeling, T helper cell differentiation, and interleukin-2 regulation, which are critical in AD pathogenesis and severity. Our findings demonstrate the utility of machine learning-integrated epigenomics in identifying robust, disease-specific biomarkers for AD diagnosis and monitoring, offering new insights into the molecular mechanisms underlying childhood AD. However, further validation in large-scale independent cohorts is required to confirm their clinical robustness and generalizability.

Big Data and Cognitive Computing doi: 10.3390/bdcc10040100

Authors: Henrique Tostes de Sousa Leo Natan Paschoal

The growing integration of chatbots and large language models (LLMs) into society raises important concerns about their potential to reproduce toxic human behaviors. As a result, it is essential to investigate these models to mitigate or eliminate such risks. This paper presents an experimental study evaluating the responses of the Llama and Alpaca models to scenarios involving verbal harassment. The methodology involved using harassment dialogues generated by an LLM as prompts to elicit responses from both models. The responses were then analyzed for levels of toxicity, sexually explicit content, and flirtatiousness. The results indicate that although both models reduce explicit offensive terms, they exhibit limitations in identifying and intercepting abusive behavior from users. Statistical analysis reveals that general-purpose instruction tuning in Alpaca does not provide a robust safety barrier compared to the Llama base model for most variables investigated in the experiment. However, a significant difference was observed concerning flirting, where Llama proved more prone to validation and encouragement than Alpaca. Furthermore, the study identifies critical vulnerabilities, such as a “self-deprecation” bias in Llama and “mirroring” behavior in Alpaca. We also report a complementary triangulation with GPT-family models as a secondary point of reference. This paper discusses and contains content that can be offensive or upsetting.

Big Data and Cognitive Computing doi: 10.3390/bdcc10030099

Authors: Oraya Sooknit Jakkarin Suksawatchon Ureerat Suksawatchon

Next Point-of-Interest (POI) recommendation focuses on predicting a user’s subsequent location based on historical check-in data. In practice, however, check-in logs frequently contain uncertain records in which ambiguous spatial, temporal, or behavioral information obscures the underlying mobility regularities, thereby degrading prediction performance. To address this challenge, this study first infers user preferences from historical trajectories and reweights transition importance based on temporal and spatial proximity. It then models transition relationships using three complementary feature dimensions: POI category, spatial area, and routine versus non-routine behavioral patterns. Using transition probability analysis, feature-level dependencies in user mobility are systematically investigated. The findings demonstrate that these transition features contribute unevenly to predictive performance, with area-based transitions yielding the strongest results when used in isolation. Nonetheless, their joint integration consistently achieves the highest accuracy, underscoring the critical role of transition-aware modeling. Across two real-world datasets, the proposed framework consistently achieves state-of-the-art performance in top-ranked accuracy (Recall@1) and ranking quality (NDCG@1), while delivering competitive effectiveness at higher cutoff values (k=3 and k=5). Notably, on the NYC dataset, MTF-POI achieves the highest Recall@1 (+19.01% over the strongest baseline) with a marginal trade-off at Recall@3, reflecting the framework’s design emphasis on precise next-step prediction.

Big Data and Cognitive Computing doi: 10.3390/bdcc10030098

Authors: Md Jahangir Alam Khondkar Ajan Ahmed Stephanie Schuckers Masudul H. Imtiaz

Speech enhancement through denoising is essential for maintaining signal intelligibility and quality in biometric speaker verification pipelines that operate in acoustically adverse conditions. Despite the proliferation of deep learning (DL) architectures for speech denoising, simultaneously optimizing noise attenuation, perceptual fidelity, and speaker-identity preservation remains an open problem. We address this gap by benchmarking three architecturally distinct DL-based enhancement models—Wave-U-Net, CMGAN, and U-Net—on three independent, domain-diverse corpora (SpEAR, VPQAD, and Clarkson) that the models never encountered during training and by introducing commercial-grade VeriSpeak speaker-verification scores as a biometric evaluation dimension absent from prior comparative studies. Our experiments reveal a clear three-way trade-off: U-Net achieves the highest signal-to-noise ratio (SNR) gains (+61.44% on SpEAR, +67.05% on VPQAD, +235.3% on Clarkson) but sacrifices naturalness; CMGAN yields the best perceptual evaluation of speech quality (PESQ) values (3.33, 1.35, and 2.50, respectively), favoring listening-comfort applications; and Wave-U-Net delivers the strongest biometric fidelity (VeriSpeak improvements of +11.63%, +30.22%, and +29.24%) while offering competitive perceptual quality. These results highlight that model selection must be driven by the target deployment scenario and provide actionable guidance for improving biometric verification robustness under real-world noise.

Big Data and Cognitive Computing doi: 10.3390/bdcc10030097

Authors: Awwal Ahmed Anthony Rispoli Carrie Wasieloski Ifrah Khurram Rafael Zamora-Resendiz Destinee Morrow Aijuan Dong Silvia Crivelli

Obstructive Sleep Apnea (OSA) is a common sleep disorder associated with serious health risks. This study leverages large language models (LLMs) to process and interpret clinical narratives in electronic health records. It develops clinically meaningful lexicons for predicting mortality and readmission risk, as well as for multiclass diagnostic classification in OSA patients. Using LLM-expanded lexicons, logistic regression models achieved ROC–AUC scores of 0.844 for 6-month all-cause post-discharge mortality, 0.817 for 1-year all-cause post-discharge mortality, and 0.729 for all-cause hospital readmissions following the first discharge. Diagnostic performance was highest with smaller n-gram representations, indicating that additional contextual length did not improve performance. Compared with frequency-based n-gram models, LLM-expanded lexicons yielded sparser feature sets with lower computational cost and comparable performance. Our findings highlight the potential of LLM-expanded lexicons to enhance OSA diagnosis and clinical risk stratification.

Big Data and Cognitive Computing doi: 10.3390/bdcc10030096

Authors: Sera Kim Youngjun Kim Jaewon Lee Dalwon Jang

This paper proposes a novel method for measuring music similarity. Existing music similarity measurements have often been used for music appreciation, but this paper proposes a method for measuring the similarity between music samples which are used for music production. Conventional music recommendation approaches often rely on either metadata-based similarity or audio-based feature similarity in isolation, which limits their effectiveness in sample-based recommendation scenarios where both compositional context and acoustic characteristics are important. To address this limitation, the proposed framework combines a hypergraph-based information similarity module with a feature-based similarity module learned using Siamese networks and triplet loss. In the information-based module, metadata attributes such as beats per minute (BPM), genre, chord, key, and instrument are modeled as vertices in a hypergraph, and Random Walk–Word2Vec embeddings are learned to capture structural relationships between music samples and their attributes. In parallel, the feature-based module employs vertex-specific Siamese networks trained on instrument and key classification tasks to learn perceptual similarity directly from audio signals. The two modules are trained independently and jointly utilized at the recommendation stage to provide attribute-specific similarity results for a given query sample. Results show that the proposed system achieves high Precision@k across multiple attributes and forms stable similarity structures in the embedding space, even without relying on user interaction data. These results reflect embedding consistency evaluated over the entire dataset where training and retrieval are performed on the same sample pool, rather than generalization to unseen samples. These results demonstrate that the proposed hybrid framework effectively captures both structural and perceptual similarity among music samples and is well suited for sample-based music recommendation in music production environments.

Big Data and Cognitive Computing doi: 10.3390/bdcc10030095

Authors: Wenxi Sheng Chi Wei Yinuo Zhang Bowen Zhang Jingyun Sun

Text-based automatic compliance checking (ACC) employs natural language processing technologies to scrutinize a corporation’s business documents, ensuring adherence to related normative texts. The current methods fall into two primary categories: symbol-based and embedding-based approaches. Symbol-based methods, noted for their accuracy and transparent processing, suffer from limited versatility. Conversely, embedding-based methods operate independently of expert knowledge yet often yield challenging-to-interpret results and require substantial volumes of annotated data. While both types of methods exhibit advantages in different aspects, the current research fails to combine these advantages effectively. Therefore, the existing methods fail to balance interpretability, generalization ability, and accuracy, which are key requirements for practical compliance systems. To address this problem, we introduce a novel approach termed the Dynamic Prompt-based Logic-Aided Compliance Checker (DPLACC), which is grounded in the prompt learning framework. This method initially parses target texts, transforming the results into first-order logical expressions. It subsequently retrieves pertinent knowledge from a knowledge graph, converting the knowledge into analogous first-order logical expressions. These expressions are then encoded into a global semantic vector via a pre-trained first-order logistic encoder. Ultimately, the semantics of expressions and initial texts are amalgamated within the prompt template, facilitating the logical knowledge enhancement of model reasoning. Experiments on Chinese and English datasets demonstrate that DPLACC comprehensively outperforms existing methods based solely on symbols or embeddings in terms of accuracy, precision, recall, and F1 score and significantly surpasses current mainstream large language models. Furthermore, DPLACC exhibits enhanced interpretability and reduced data dependence, maintaining 70% checking accuracy with as few as ten training samples. This capability allows DPLACC to be rapidly deployed in data-scarce real-world scenarios with minimal annotation overhead, thus offering a practical pathway toward the scalable implementation of compliance inspection systems.

Big Data and Cognitive Computing doi: 10.3390/bdcc10030094

Authors: Abdussalam Elhanashi Siham Essahraui Pierpaolo Dini Davide Paolini Qinghe Zheng Sergio Saponara

Generative artificial intelligence and foundation models have changed machine learning by allowing systems to produce readable text, realistic images, and other multimodal content with little direct input from a user. Foundation models are large neural networks trained on very large and varied datasets, and they form the core of many current generative AI (GenAI) systems. Their rapid development has led to major advances in areas like natural language processing, computer vision, multimodal learning, and robotics. Examples include GPT, LLaMA, and diffusion-based architectures, such as models often used for image generation. Systems such as Stable Diffusion show this shift by illustrating how AI can interpret information, draw basic inferences, and produce new outputs using more than one type of data. This review surveys common foundation model architectures and examines what they can do in generative tasks. It reviews Transformer, diffusion, and multimodal architectures, focusing on methods that support scaling and transfer across domains. The paper also reviews key approaches to pretraining and fine-tuning, including self-supervised learning, instruction tuning, and parameter-efficient adaptation, which support these systems’ ability to generalize across tasks. In addition to the technical details, this review discusses how GenAI is being used for text generation, image synthesis, robotics, and biomedical research. The study also notes continuing challenges, such as the high computing and energy demands of large models, ethical concerns about data bias and misinformation, and worries about privacy, reliability, and responsible use of AI in real settings. This review brings together ideas about model design, training methods, and social implications to point future research toward GenAI systems that are efficient, easy to interpret, and reliable, while supporting scientific progress and ethical responsibility.

Big Data and Cognitive Computing doi: 10.3390/bdcc10030093

Authors: Phumudzo Lloyd Seabe Claude Rodrigue Bambe Moutsinga Maggie Aphane

Accurate forecasting of equity returns remains fundamentally constrained by weak short-horizon predictability, pronounced noise, and structural non-stationarity. While deep learning models have been widely applied to financial time series, most studies prioritize point prediction and provide limited guidance on reliable uncertainty quantification, particularly in emerging markets. This study developed an uncertainty-aware forecasting framework for the South African equity market by integrating variational mode decomposition (VMD), gated recurrent units (GRUs), and temporal conformal prediction (TCP) to construct distribution-free prediction intervals with finite-sample coverage guarantees. Using daily returns from the FTSE/JSE All Share Index, we first confirmed that baseline recurrent models applied directly to raw returns exhibited negligible out-of-sample explanatory power, consistent with weak-form market efficiency. Incorporating VMD enhanced representation learning and improved point forecast accuracy by isolating latent frequency components. However, model-based predictive variance alone proved insufficient for reliable calibration. Embedding the models within a rolling conformal prediction framework restored near-nominal coverage across multiple confidence levels while allowing interval widths to adapt dynamically to changing volatility regimes. Robustness analyses, including walk-forward validation, stress-regime evaluation, and block permutation negative control experiments, indicated that the observed performance was not driven by temporal leakage or alignment artifacts. The results further highlight a trade-off between interval sharpness and tail-risk protection, particularly during extreme market events. Overall, the findings support a shift from return-level prediction toward calibrated uncertainty estimation as a more stable and economically meaningful objective in non-stationary financial environments.

Big Data and Cognitive Computing doi: 10.3390/bdcc10030092

Authors: Bobur Saidov Vladimir Barakhnin Rakhmon Saparbaev Zayniddin Narmuratov Rustamova Manzura Ruzmetova Zilolakhon Anorgul Atajanova

This work proposes a hybrid Uzbek sentiment analysis model (sometimes referred to as tonality analysis in the local literature) that integrates contextual text representations with named-entity information from an NER module and emoji-based emotional cues that are common in short online messages. To provide a comprehensive baseline comparison, we evaluate seven approaches—SVM, LSTM, mBERT, XLM-RoBERTa-base, mDeBERTa-v3, LaBSE, and the proposed hybrid model—covering both classical machine learning and modern multilingual transformer architectures for low-resource sentiment tasks. The overall pipeline begins with Uzbek-specific text normalization to reduce noise from informal spellings, transliteration variants, and inconsistent apostrophe usage. In parallel, the system performs explicit emoji extraction to capture affective signals that are often expressed non-verbally in social media texts. Next, we construct three complementary feature streams: a context encoder for sentence-level semantics, NER-driven entity features that encode entity mentions and types, and an emotion module that models emoji priors and their interaction with contextual meaning. These streams are fused into a unified representation and fed to a final classifier to predict sentiment polarity. Experiments on an Uzbek test set demonstrate that the hybrid model reaches an F1-score of 0.92, consistently outperforming text-only baselines. The results indicate that entity-aware and emoji-informed features improve robustness under sarcasm/irony, mixed sentiment with multiple targets, and orthographic noise, making the approach suitable for social media analytics, public opinion monitoring, customer feedback triage, and recommendation-oriented text mining.

Big Data and Cognitive Computing doi: 10.3390/bdcc10030091

Authors: Gang Cheng Hailin Pei Xiaokang Chen Xiaorong Pang Renzheng Sun

Spontaneous combustion in goaf areas poses a significant threat to coal mine safety. Traditional safety management systems, reliant on passive response and single-indicator thresholds, often suffer from delayed warnings and lack cognitive decision support. To address this challenge, this study proposes a big-data-driven cognitive computing framework for dynamic risk prediction of goaf spontaneous combustion, based on a “Cloud-Edge-End” collaborative architecture. The method leverages multi-sensor big data streams (CO, C2H4, O2, etc.) and deploys a lightweight Radial Basis Function (RBF) neural network on underground edge computing nodes (STM32) for real-time analytics. The model demonstrates excellent predictive performance on imbalanced datasets, with a PR-AUC of 0.910 and a recall of 99.7%. The edge-deployed RBF model achieves a single-pass inference time of only 0.62 ms, enabling real-time cognitive risk mapping. Field application at Z Coal Mine validated the system’s effectiveness, providing an average pre-warning time of 48.5 h, achieving zero spontaneous combustion accidents, and reducing the Total Recordable Injury Rate (TRIR) by 15.2%. This work illustrates how edge-based cognitive computing can transform safety management from passive response to proactive prevention, offering a scalable and interpretable framework for intelligent mine safety.

Big Data and Cognitive Computing doi: 10.3390/bdcc10030090

Authors: Abheek Pradhan Sana Alamgeer Rakesh Suvvari Syed Tousiful Haque Anne H. H. Ngu

Wearable fall detection systems face a fundamental challenge: while gyroscope data provide valuable orientation cues, naively combining raw gyroscope and accelerometer signals can degrade performance due to noise contamination. To overcome this challenge, we present a dual-stream transformer architecture that incorporates (i) Kalman-based sensor fusion to convert noisy gyroscope angular velocities into stable orientation estimates (roll, pitch, yaw), maintaining an internal state of body pose, and (ii) processing accelerometer and orientation streams in separate encoder pathways before fusion to prevent cross-modal interference. Our architecture further integrates Squeeze-and-Excitation channel attention and Temporal Attention Pooling to focus on fall-critical temporal patterns. Evaluated on the SmartFallMM dataset using 21-fold leave-one-subject-out cross-validation, the dual-stream Kalman transformer achieves 91.10% F1, outperforming single-stream Kalman transformers (89.80% F1) by 1.30% and single-stream baseline transformers (88.96% F1) by 2.14%. We further evaluate the model in real time using a watch-based SmartFall App on five participants, maintaining an average F1 score of 83% and an accuracy of 90%. These results indicate robust performance in both offline and real-world deployment settings, establishing a new state-of-the-art for inertial-measurement-unit-based fall detection on commodity smartwatch devices.

Big Data and Cognitive Computing doi: 10.3390/bdcc10030089

Authors: Elena Ryumina Alexandr Axyonov Mikhail Dolgushin Dmitry Ryumin Alexey Karpov

Automated video-based detection of cognitive disorders can enable a scalable non-invasive health monitoring. However, existing methods focus on a single disease and provide limited interpretability, whereas real-world videos often contain co-occurring conditions. We propose a novel unified multi-task method to detect depression and Parkinson’s disease (PD) from in-the-wild video data called DEPART (DEpression and PArkinson’s Recognition Technique). It performs body region extraction, Contrastive Language-Image Pre-training (CLIP)-based visual encoding, Transformer-based temporal modeling, and prototype-aware classification with a gated fusion technique. Gradient-based attention maps are used to visualize task-specific regions that drive predictions. Experiments on the In-the-Wild Speech Medical (WSM) corpus demonstrate competitive performance: the multi-task model achieves Recall of 82.39% for depression and 78.20% for PD, compared with 87.76% and 78.20%, for the best single-task models. The multi-task learning initially increases false positives for healthy persons in the PD subset, mainly due to annotation–modality mismatches, static visual content misinterpreted as motor impairments, and occasional body detection failures. After cleaning the test data, Recall for healthy individuals becomes comparable across models; the multi-task model improves Recall for both depression (from 82.39% to 87.50%) and PD (from 78.20% to 86.14%), suggesting better robustness for real-life clinical applications.

Big Data and Cognitive Computing doi: 10.3390/bdcc10030088

Authors: Saule Kudubayeva Yernar Seksenbayev Aigerim Yerimbetova Elmira Daiyrbayeva Bakzhan Sakenov Duman Telman Mussa Turdalyuly

Multimodal human–AI systems generally consider facial expressions and body motions as separate input streams, leading to disjointed interpretations and diminished emotional coherence. To overcome this issue, we offer the Engagement-Safe Expressive Alignment (ESEA) paradigm and the Unified Visual Synchrony (UVS) framework as its computational implementation. UVS models the coherence between facial expressions and gestures, offering an interpretable visual synchrony signal that can function as adaptive feedback in human–AI interactions. The framework’s key component is the Consistency Index for Affective Synchrony (CIAS), which correlates brief visual segments with scalar synchrony scores through a common latent representation. Facial and gestural signals are processed by modality-specific projection networks into a unified latent space, and CIAS is derived from the similarity and short-term temporal consistency of these latent trajectories. The synchrony index is regarded as an estimation of affective visual coherence within the ESEA paradigm. We formalize the UVS/CIAS framework and conduct a comparative experimental evaluation utilizing matched and mismatched face–gesture segments derived from rendered dialog footage. Utilizing ROC analysis, score distribution comparisons, temporal visualizations, and negative control tests, we illustrate that CIAS effectively captures structured face–gesture alignment that surpasses similarity-based baselines, while also delivering a persistent, time-resolved synchronization signal. These findings establish CIAS as a principled and interpretable feedback signal for future affect-aware, engagement-focused multimodal agents.

Big Data and Cognitive Computing doi: 10.3390/bdcc10030087

Authors: Junyi Jiang Dong Li Qingyi Yang Zhenhua Zhang Lei Wang Wenru Zhao Mingliang Chen

Slope instability can cause severe disasters, making stability prediction essential. Machine learning has become a key tool for this purpose, as it avoids complex mechanical calculations and efficiently handles high-dimensional data. Currently, the data used in machine learning primarily originate from real-world cases. However, such cases are inherently limited in quantity and often fail to comprehensively represent all potential slope conditions. To address these limitations, this study proposes a method for constructing numerical simulation databases. Based on this, we develop a model establishment method for rapid evaluation of slope stability integrating numerical simulation with engineering cases. This study uses six characteristic parameters to assess slope stability, including unit weight γ, cohesion c, internal friction angle φ, slope angle α, slope height H, and pore pressure ratio ru. Through extensive literature mining, we established a database of 684 engineering cases. Based on statistical analysis of input parameters, a numerical simulation scheme was designed. Batch calculations were performed using MATLAB to determine simulation results. The engineering case database was then partitioned into training and testing sets for model development and validation. Subsequently, the numerical simulation database was incorporated into the training set for retesting. Results demonstrate that when considering all predictive indicators, the prediction accuracy of the GRNN-based model improved from 85% to 88.3%, while the PNN-based model showed an increase from 69% to 88.3%. This study offers new insights for optimizing numerical simulation design and enhancing machine learning performance in slope stability prediction.

Big Data and Cognitive Computing doi: 10.3390/bdcc10030086

Authors: Langgao Cheng Yanying Mao Guowang Li Honghui Chen

Generative recommendation systems based on Large Language Models leverage their reasoning capabilities to capture users’ latent interests. However, aligning continuous user behavioral embeddings with the discrete semantic space of LLMs remains a challenge. Direct alignment often leads to semantic mismatch and hallucination issues. Furthermore, existing methods typically rely on multi-stage training strategies to adapt to variations in feature distributions, thereby limiting training efficiency. To address the aforementioned issues, we propose SBT-Rec, a structured behavioral tokenization framework. Specifically, we first design a hierarchical discrete structure discovery module, utilizing a recursive residual quantization mechanism to decompose continuous behavioral vectors into discrete behavioral atoms to resolve modality discrepancies. Second, the multi-scale behavioral semantic reconstruction module reconstructs behavioral representations via residual superposition, thereby reducing data noise. Third, a residual-aware modality distribution aligner is introduced to transform behavioral features into input tokens compatible with the LLM via non-linear mapping. Finally, based on structured discrete representations, we propose a single-stage behavioral-semantic adaptive optimization strategy, achieving end-to-end parameter-efficient fine-tuning. Experiments on the MovieLens, LastFM, and Steam datasets demonstrate that SBT-Rec outperforms existing baseline models in terms of recommendation accuracy, training efficiency, and noise robustness.

Big Data and Cognitive Computing doi: 10.3390/bdcc10030085

Authors: Patrick Steinert Stefan Wagenpfeil Ingo Frommholz Matthias L. Hemmje

After the peak of the recent hype wave of interest surrounding the metaverse, virtual world applications remained in areas such as gaming, VR training, simulations, and collaboration. In this context, recordings are created which subsequently evolve into extensive collections that users may wish to access, search through, and retrieve items from. In order to facilitate searchability of metaverse recordings, it is necessary to adapt content analysis and indexing techniques to the specific characteristics of these recordings. This paper presents a reference model, the Processing Framework for Metaverse Recordings (PFMR), which details the phases of structural analysis, feature extraction, data mining, and feature fusion. The objective is to facilitate efficient retrieval of metaverse content. Our evaluation, based on a prototypical implementation, demonstrates the applicability and effectiveness of PFMR. This lays the groundwork for further integration of metaverse-specific content into Multimedia Information Retrieval systems. The evaluation of the 256 Metaverse Recording dataset shows that PFMRs’ domain-specific adaptability and integratability allows effective metaverse recording information retrieval for metaverse-specific features such as avatar detection, dialog mining, and toxicity classification.

Big Data and Cognitive Computing doi: 10.3390/bdcc10030084

Authors: Zoltán Mészáros M. A. Hannan Bin Azhar Tasmina Islam Soumya Kanti Manna

Home based telerehabilitation has expanded after COVID-19, but delivering timely guidance and monitoring exercise performance outside the clinic remains difficult. Traditional physiotherapy often relies on repeated execution of simple routines, yet clinicians have limited visibility into adherence and movement quality during unsupervised sessions. From a systems perspective, many telerehabilitation approaches also face constraints in accessibility, bandwidth, and computational cost that can limit practical deployment. This paper presents a modular telerehabilitation framework and prototype that captures and records rehabilitation exercise sessions for asynchronous clinician review in a 3D visualisation environment. The system integrates skeletal motion capture with plantar pressure sensing, and stores sessions as portable artefacts to support replay, inspection, and downstream analysis. A connector-based architecture enables extension to additional sensors without redesigning the core application, and the design aims to support deployment under constrained home computing and networking conditions. The manuscript contributes an implementation blueprint and reference architecture for multimodal capture and replay. Clinical effectiveness, usability outcomes, and quantitative sensor accuracy benchmarking are outside the scope of this work and are identified as necessary future evaluation.

Big Data and Cognitive Computing doi: 10.3390/bdcc10030083

Authors: Giuseppe Ciaburro Virginia Puyana-Romero

Sound Event Detection (SED) is a growing area with vast prospects for understanding and designing the sonic fabric of smart cities. In this paper, the latest advances in SED are summarized, focusing on models, datasets, and applications from scientific papers listed on Scopus and Web of Science. The paper provides a clear view of how SED is being used in smart cities, public safety, environment monitoring, and home security. The paper also addresses the challenges of SED, including dataset representativeness, model robustness under noisy or complex acoustic scenes, event rarity detection, as well as the ethics of using automatic listening. The paper also provides a view of future work to be undertaken in SED. The focus of the paper is on self-supervised learning, multi-modal fusion, neuro-inspired approaches, as well as privacy-preserving analytics. The paper provides a view of SED as a key technology to make smart cities safe, secure, and sustainable. SED has vast prospects as a key technology to enable artificial perception of smart cities.

Big Data and Cognitive Computing doi: 10.3390/bdcc10030082

Authors: Xiyun Yang Han Chen Xiangjun Li Xiaoyu Liu

Accurate carbon price forecasting is vital for risk management but is hindered by high volatility and sensitivity to external shocks. Existing multivariate models typically overlook unstructured news sentiment, failing to capture irrational fluctuations driven by market public opinion. To address this, this paper proposes VBN-Net, a hybrid model integrating carbon-specific news sentiment with Variational Mode Decomposition (VMD). Two core innovations are presented: First, a multi-modal input mechanism combines structured financial data with unstructured carbon news sentiment to effectively capture policy-driven shocks. Second, a Sequential Beluga Whale Optimization strategy is designed to adaptively optimize feature engineering in steps. Unlike conventional approaches, the VBN-Net first employs VMD for denoising and frequency decomposition, and then optimizes the fusion weights of news sentiment across different frequency components derived from multi-source news. This strategy effectively overcomes the subjectivity of manual parameter selection, providing high-quality features for a fixed CNN-BiLSTM backbone. By integrating VMD-based denoising with optimized multi-source news fusion, the model achieves consistent performance improvements across multiple evaluation metrics. The empirical findings validate the effectiveness of the proposed model in enhancing forecasting performance, thereby providing a reliable analytical tool for participants in the carbon market.

Big Data and Cognitive Computing doi: 10.3390/bdcc10030081

Authors: Chongwen Tian Rong Li

This study proposes the Double-Ensemble Learning Classification with SMOTE (DELC-SMOTE), a novel hierarchical framework designed to address the critical challenge of severe class imbalance in financial bond default prediction. The model integrates the Synthetic Minority Over-sampling Technique (SMOTE) into a two-phase ensemble architecture. The first phase employs introspective stacking, where six heterogeneous base learners are individually enhanced through algorithm-specific balancing and meta-learning. The second phase fuses these optimized experts via performance-weighted voting. Empirical analysis utilizes a comprehensive dataset of 10,440 Chinese corporate bonds (522 defaults, ~5% default rate) sourced from Wind and CSMAR databases. Given the high cost of both false negatives and false positives in risk assessment, the Geometric Mean (G-mean) and Specificity are employed as primary evaluation metrics. Results demonstrate that the proposed DELC-SMOTE model significantly outperforms individual base classifiers and benchmark ensemble variants, achieving a G-mean of 0.9152 and a Specificity of 0.8715 under the primary experimental setting. The model exhibits robust performance across varying imbalance ratios (2%, 10%, 20%) and strong resilience against data noise, perturbations, and outliers. These findings indicate that the synergistic integration of data-level resampling within a diversified, two-tiered ensemble structure effectively mitigates class imbalance bias and enhances predictive reliability. The framework offers a robust and generalizable tool for actionable default risk assessment in imbalanced financial datasets.

Big Data and Cognitive Computing doi: 10.3390/bdcc10030080

Authors: Abu Anas Ibn Samad Tanvir Ahmed Md Nazmul Huda

Intelligent Transportation Systems (ITS) primarily rely on flow rate and occupancy to estimate traffic states. However, in heterogeneous traffic conditions characterized by weak lane discipline and diverse vehicle classes, these conventional metrics fail to capture the true operational efficiency of signalized intersections. High flow rates can mask underlying inefficiencies, while low flow rates do not necessarily indicate free-flow conditions. This paper introduces a novel computer vision-based metric, the Effective Flow Ratio (EFR), designed to quantify the actual discharge efficiency of mixed traffic. By leveraging Bird’s-Eye View (BEV) vehicle tracking using You Only Look Once version 11 (YOLOv11) and ByteTrack, EFR distinguishes between kinematic movement and effective discharge, resolving the ambiguity of “moving but not clearing” states. We analyze 21 days of continuous footage from a rooftop-mounted camera overlooking a congested intersection in Dhaka, Bangladesh, exhibiting distinct non-linear behaviors compared to raw flow counts. Our results demonstrate that: (i) Flow rate and discharge efficiency are dynamically decoupled, evidenced by significant variance in EFR within identical flow bins; (ii) Temporal rolling correlations reveal transient regimes where traditional signal control logic would misinterpret congestion severity; and (iii) EFR provides a more robust proxy for intersection performance than occupancy or volume alone. The proposed metric offers a granular, physics-informed input for next-generation adaptive traffic signal control in developing urban environments.

Big Data and Cognitive Computing doi: 10.3390/bdcc10030079

Authors: Volodymyr Dubetskyy Maria-Dolores Cano

Fully Homomorphic Encryption (FHE) provides strong data confidentiality during computation but often suffers from high latency on Central Processing Units (CPUs). This study evaluates Graphics Processing Unit (GPU) acceleration for modern FHE libraries across a laptop (NVIDIA GTX 1650 Ti), a server (NVIDIA RTX 4060), and a Jetson Nano 2 GB embedded GPU. We benchmark key generation, arithmetic operations, Boolean-gate evaluation and scheme-specific tasks such as relinearization and key switching, using library-provided benchmarks with an explicit baseline (operation scope, timing boundaries, and parameter tuples). Moreover, we compare GPU-native libraries (NuFHE, Phantom-FHE, and Troy-Nova) with CPU-oriented ones (Microsoft SEAL, HElib, OpenFHE, Cupcake, and TFHE-rs). Results show GPUs deliver significant speedups for targeted operations. For example, NuFHE’s NVIDIA CUDA (Compute Unified Device Architecture) backend achieves about 1.4× faster Boolean-gate evaluation on the laptop and 3.4× faster on the server compared to its OpenCL backend. Likewise, RLWE (Ring Learning With Errors)-based schemes (BFV, CKKS, and BGV) see marked gains for polynomial arithmetic such as Number Theoretic Transform (NTT) when executed via Phantom-FHE. However, attempts to add CUDA support to Microsoft SEAL reveal four main challenges: high-precision modular arithmetic on GPUs, sequential dependencies in SEAL’s design, limited GPU memory and complex build-system changes. In light of these findings, we propose revised guidelines for GPU-first FHE libraries and practical recommendations for deploying high-throughput, privacy-preserving solutions on modern GPUs.

Big Data and Cognitive Computing doi: 10.3390/bdcc10030078

Authors: M. Hadi Sepanj Benyamin Ghojogh Saed Moradi Paul Fieguth

Self-supervised learning (SSL) has emerged as a powerful paradigm for representation learning by optimizing geometric objectives, such as invariance to augmentations, variance preservation, and feature decorrelation, without requiring labels. However, most existing methods operate in Euclidean space, limiting their ability to capture nonlinear dependencies and geometric structures. In this work, we propose Kernel VICReg, a novel self-supervised learning framework that pulls the VICReg objective into a Reproducing Kernel Hilbert Space (RKHS). By kernelizing each term of the loss, variance, invariance, and covariance, we obtain a general formulation that operates on double-centered kernel matrices and Hilbert–Schmidt norms, enabling nonlinear feature learning without explicit mappings. We demonstrate that Kernel VICReg mitigates the risk of representational collapse under challenging conditions and improves performance on datasets exhibiting nonlinear structure or limited sample regimes. Empirical evaluations across MNIST, CIFAR-10, STL-10, TinyImageNet, and ImageNet100 show consistent gains over Euclidean VICReg, with particularly strong improvements on datasets where nonlinear structures are prominent. UMAP visualizations are provided only as a qualitative illustration of embedding geometry and are not used as a calibration or statistical validation. Our results suggest that kernelizing SSL objectives is a promising direction for bridging classical kernel methods with modern representation learning.

Big Data and Cognitive Computing doi: 10.3390/bdcc10030077

Authors: Manlika Seefong Panuwat Wisutwattanasak Kestsirin Theerathitichaipa Pattarawadee Prasomsab Nisa Dackuntod Thanapong Champahom Rattanaporn Kasemsri

Hospital transport services represent a vital alternative for addressing inequities in access to medical care, particularly in countries where public transportation systems are inadequate, such as Thailand. This approach enables equitable and widespread access to healthcare services for residents in underserved areas. The objective of this study is to analyze the factors influencing the choice of hospital transport travel mode by comparing various machine learning algorithms. The findings reveal that the categorical boosting model outperformed the other models across all performance metrics. The model results indicate that waiting time, travel time, travel cost, and comfortability significantly influence the decision to use hospital transport services. Furthermore, demographic data analysis highlights critical factors such as age, gender, income, travel frequency, occupation, and time of travel, all of which significantly affect the choice of hospital transport service. To maximize the practical implications of this study, policy recommendations and implementation strategies are proposed to support decision-makers in promoting equitable travel options and eliminating barriers to fair access to healthcare services.

Big Data and Cognitive Computing doi: 10.3390/bdcc10030076

Authors: Aminu Musa Rajesh Prasad Peter Onwualu Monica Hernandez

Deep learning has achieved expert-level performance in medical imaging analysis. However, models often fail to generalize across patient populations due to cross-population domain shifts, distributional differences arising from demographic variability, variations in imaging protocols, scanner hardware, and differences in disease prevalence. This challenge limits the real-world deployment and can increase health inequities. This review systematically examines the nature, causes, and impact of cross-population domain shift in deep learning-based medical imaging analysis. We analyzed 50 peer-reviewed studies from 2020 to 2025, evaluating the proposed methodologies for handling population shifts, the datasets employed, and the metrics used to assess performance. Our findings demonstrate that performance degradation ranged from 10–25% when models were tested on unseen populations, emphasizing the substantial impact of domain shifts on model generalizability. The literature reveals that mitigation strategies broadly fall into two categories: data-centric approaches, such as augmentation and harmonization, and model-centric approaches, including domain adaptation, transfer learning, adversarial learning, multi-task learning, and continual learning. While domain adaptation and transfer learning are the most widely used, their performance gains across populations remain modest, ranging from 5–15%, and are not supported by external validation. Our synthesis reveals a significant reliance on large, publicly available datasets from limited regions, with an underrepresentation of data from low- and middle-income countries. Evaluation practices are inconsistent, with few studies employing standardized external test sets. This review provides a structured taxonomy of mitigation techniques, a refined analysis of domain shift characteristics, and an in-depth critique of methodological challenges. We highlight the urgent need for more geographically and demographically inclusive datasets, adaptable modeling techniques, and standardized evaluation protocols to enable accurate and equitable AI-driven diagnostics across diverse populations. Finally, we outline future research directions to guide the development of robust, generalizable, and fair models for medical imaging analysis.

Big Data and Cognitive Computing doi: 10.3390/bdcc10030075

Authors: Tanmay Sunil Hatkar Abhinav Pandey Saad B. Ahmed

Semantic segmentation plays a pivotal role in autonomous driving, enabling pixel-level understanding of road scenes. Although transformer-based models such as SegFormer have shown exceptional performance on large datasets, their generalization to smaller and geographically diverse datasets remains underexplored. In this work, we analyze the scalability and transferability of SegFormer variants (B3, B4, B5) using CamVid as the base dataset. We perform cross-dataset transfer learning to KITTI and IDD, evaluate class-level performance, and explore explainable AI via confidence heatmaps. Our findings show that SegFormer-B5 achieves the highest accuracy (82.4% mIoU) on CamVid, while transfer learning from CamVid improves mIoU on KITTI by 2.57% and enhances class-specific predictions in IDD by over 70%. These results highlight the practical potential of SegFormer in real-world segmentation systems and the interpretability benefits of confidence-based visual analysis.

Big Data and Cognitive Computing doi: 10.3390/bdcc10030074

Authors: Nikitas Gerolimos Vasileios Alevizos Georgios Priniotakis

Ergonomic load in human–autonomy teams is commonly treated as a static score or a post-hoc audit, even though modern sensing and communication enable real-time regulation of operator effort. We model ergonomic load as a dissipative dynamical state inferred online from multimodal effort proxies and task context, and couple it to autonomy through load-dependent gain moderation and compliance shaping. The method is evaluated on public human–swarm and human–robot interaction traces together with effort-proximal wearable and myographic datasets using a unified, windowed pipeline and controlled stress tests that emulate latency, downsampling, packet loss, and channel dropouts. On a large human–swarm benchmark, the estimator achieves strong discrimination and calibration for rare high-load events (up to AUROC 0.87, AUPRC 0.41, ECE 0.031 at q=0.90) and degrades predictably under delay, with a knee around 300–400ms (AUROC 0.87→0.80, ECE 0.031→0.061 at 500ms). Embedding the estimate in the adaptation schedule reduces overload incidence and oscillatory redistribution while preserving coordination proxies in surrogate closed-loop simulation: overload time drops from 7.8% to 4.1% (relative reduction ≈ 47%) with throughput maintained near baseline (1.00→0.97) and oscillation power reduced (0.26→0.14) under nominal timing. These results provide a reproducible pathway for making ergonomics a control-relevant feedback signal, together with explicit operational constraints on estimator calibration (target ECE ≤0.05) and end-to-end latency (effective τ≤300ms) required to avoid regime switching and maintain stable, interpretable adaptation.