Search Results (139)

Background and Objectives: Glaucoma (GL) classification is crucial for early diagnosis and treatment, yet relying solely on stand-alone models or International Classification of Diseases (ICD) codes is insufficient due to limited predictive power and inconsistencies in clinical labeling. This study aims to improve GL classification using stacked weight-based machine learning models. Materials and Methods: We analyzed a subset of 33,636 participants (58% female) with 340,444 visual fields (VFs) from the Mass Eye and Ear (MEE) dataset. Five clinically relevant GL detection models (LoGTS, UKGTS, Kang, HAP2_part1, and Foster) were selected to serve as base models. Two multi-layer perceptron (MLP) models were trained using 52 total deviation (TD) and pattern deviation (PD) values from Humphrey field analyzer (HFA) 24-2 VF tests, along with four clinical variables (age, gender, follow-up time, and race) to extract model weights. These weights were then utilized to train three meta-learners, including logistic regression (LR), extreme gradient boosting (XGB), and MLP, to classify cases as GL or non-GL. Results: The MLP meta-learner achieved the highest performance, with an accuracy of 96.43%, an F-score of 96.01%, and an AUC of 97.96%, while also demonstrating the lowest prediction uncertainty (0.08 ± 0.13). XGB followed with 92.86% accuracy, a 92.31% F-score, and a 96.10% AUC. LR had the lowest performance, with 89.29% accuracy, an 86.96% F-score, and a 94.81% AUC, as well as the highest uncertainty (0.58 ± 0.07). Permutation importance analysis revealed that the superior temporal sector was the most influential VF feature, with importance scores of 0.08 in Kang’s and 0.04 in HAP2_part1 models. Among clinical variables, age was the strongest contributor (score = 0.3). Conclusions: The meta-learner outperformed stand-alone models in GL classification, achieving an accuracy improvement of 8.92% over the best-performing stand-alone model (LoGTS with 87.51%), offering a valuable tool for automated glaucoma detection. Full article

In response to the increasing demands for energy conservation and pollution reduction, optimizing transformer design to reduce operational losses and minimize raw material usage has become crucial. This paper introduces an innovative methodology that combines ensemble learning models with hybrid multi-objective optimization heuristic algorithms to optimize leakage impedance deviation, on-load loss, and raw material consumption in power transformers. The stacking ensemble model uses support vector machines, linear regression, decision tree regression, and K-nearest neighbors as base learners, with the extreme learning machine serving as the meta-learner to re-learn outputs from first-level learners. Given the significant impact of hyperparameters on the prediction performance of ensemble learning models, an improved particle swarm optimization method is proposed for effective hyperparameter optimization. To assess the uncertainty of the proposed ensemble learning model, a Kriging surrogate model-based analysis is outlined. Moreover, a powerful multi-objective algorithm that integrates the multi-objective grey wolf optimization (MOGWO) and the non-dominated sorting genetic algorithm-III (NSGA3) is presented for model optimization. This approach demonstrates superior performance compared to mainstream multi-objective optimization algorithms. The effectiveness of this method is further validated through the engineering tests of two real engineering cases. The proposed algorithm can accommodate various design requirements and, under the given constraints, achieve a multi-objective optimization design for power transformers, ensuring optimal performance in different operational scenarios. Full article

Mobile technology, especially mobile learning, has long been an emerging and thriving field, and remains a main theme in mobile learning applications and systems. The extensive utilization of mobile learning has prompted the invention of many mobile applications. As a result of rapid advances in application technologies, various learning applications can combine different media or multi-modalities, such as video, audio, images, animated graphics, and text, to create multimedia learning resources that engage learners. However, the most favorable modalities in different learning applications that assist performance are worth exploring. This study employed mixed methods to investigate the current multi-modality situation in learning application utilization among 300 university students in China, where a rapid educational technology revolution is occurring. The findings revealed that the verbal modality (M = 3.99, S*D = 0.79) and the writing modality (M = 3.99, S*D = 0.75) in the learning applications were less enjoyable and less effective at enhancing learning performance. In exam-based or function-based apps, all five modalities in this research were considered important, especially the visual and aural modes. The results of this study also revealed that a majority of university learners were satisfied with the multi-modalities in different types of applications, except for game-based apps, that assist their learning performance (56.7%, M = 3.87, S*D = 0.79), which contrasts with the results of several related studies. Overall, college users perceived that multi-modalities were effective in helping them to complete tasks, and all modalities in current applications satisfied most of the users’ needs to assist their learning performance. In the end, the findings indicated a positive and strong linear relationship [r = 0.766, p < 0.05] between multi-modalities and assisted learning performance with the help of more capable (knowledgeable) others with the use of mobile applications. Full article

Modern education faces persistent challenges, including disengagement, inequitable access to learning resources, and the lack of personalized instruction, particularly in virtual environments. In this perspective, we envision a transformative Metaverse classroom model, the Multi-layered Immersive Learning Environment (Meta-MILE) to address these critical issues. The Meta-MILE framework integrates essential components such as immersive infrastructure, personalized interactions, social collaboration, and advanced assessment techniques to enhance student engagement and inclusivity. By leveraging three-dimensional (3D) virtual environments, artificial intelligence (AI)-driven personalization, gamified learning pathways, and scenario-based evaluations, the Meta-MILE model offers tailored learning experiences that traditional virtual classrooms often struggle to achieve. Acknowledging potential challenges such as accessibility, infrastructure demands, and data security, the study proposed practical strategies to ensure equitable access and safe interactions within the Metaverse. Empirical findings from our pilot experiment demonstrated the framework’s effectiveness in improving engagement and skill acquisition, with broader implications for educational policy and competency-based, experiential learning approaches. Looking ahead, we advocate for ongoing research to validate long-term learning outcomes and technological advancements to make immersive learning more accessible and secure. Our perspective underscores the transformative potential of the Metaverse classroom in shaping inclusive, future-ready educational environments capable of meeting the diverse needs of learners worldwide. Full article

Over the last ten years, continuous attention has been paid to the use of digital games in vocabulary learning. Their effectiveness and availability have been widely discussed. However, the experiences of language learners and the underlying patterns of their engagement while using digital games for vocabulary learning remain underexplored. In order to fill this significant gap, this study aimed to examine the influential mechanism of English as a Foreign Language (EFL) learners’ flow experiences in digital game-based vocabulary learning (DGBVL). The sample consisted of 306 Chinese EFL learners who had DGBVL app usage experience, and data collection was based on a DGBVL flow experience instrument employed through an online platform. Structural equation modeling (SEM) was employed to assess the reliability and validation of the existing scale for various DGBVL apps. A multi-group analysis was then conducted, revealing that the influential mechanism was a process in which the effects of antecedents on outcomes could be mediated by flow experiences. In addition, the role of usage frequency was also explored, and three paths were found to differ across three usage frequency levels (i.e., seldom, sometimes, and always): the effect of balance of skill and challenge on enjoyment, the effect of enjoyment on satisfaction, and the effect of perceived learning on satisfaction. These findings provide new insights for the influential mechanism of flow experiences and will assist EFL learners in optimizing their learning outcomes in digital game-based vocabulary learning. Full article

As the climate emergency escalates, the role of forests in carbon sequestration is paramount. This paper proposes a framework that integrates local capacities, multi-source remote sensing data, and meta-learning to enhance forest carbon assessment methodologies in data-scarce regions. By integrating multi-source optical and radar remote sensing data alongside community forest inventories, we applied a meta-modelling approach using stacked generalization ensemble to estimate forest above-ground carbon (AGC). We also conducted a Kruskal–Wallis test to determine significant differences in AGC among different tree species. The Kruskal–Wallis test (p = 1.37 × 10⁻¹³) and Dunn post-hoc analysis revealed significant differences in carbon stock potential among tree species, with Afzelia quanzensis ($\tilde{x}$= 12 kg/ha, P-holm-adj. = 0.05) and the locally known species M’buta ($\tilde{x}$ = 6 kg/ha, P-holm-adj. = 5.45 × 10⁻⁹) exhibiting a significantly higher median AGC. Our results further showed that combining optical and radar remote sensing data substantially improved prediction accuracy compared to single-source remote sensing data. To improve forest carbon assessment, we employed stacked generalization, combining multiple machine learning algorithms to leverage their complementary strengths and address individual limitations. This ensemble approach yielded more robust estimates than conventional methods. Notably, a stacking ensemble of support vector machines and random forest achieved the highest accuracy (R² = 0.84, RMSE = 1.36), followed by an ensemble of all base learners (R² = 0.83, RMSE = 1.39). Additionally, our results demonstrate that factors such as the diversity of base learners and the sensitivity of meta-leaners to optimization can influence stacking performance. Full article

While the pet market is continuously rapidly increasing in Korea, pet dog owners feel uncomfortable in coping with pet dog’s health problems in time. In this paper, we propose a pre-diagnosis system based on neuro-fuzzy learning, enabling non-expert users to monitor their pets’ health by inputting observed symptoms. To develop such a system, we form a disease–symptom database based on several textbooks with veterinarians’ guidance and filtering. The system offers likely disease predictions and recommended coping strategies based on fuzzy inference. We evaluated three fuzzy inference algorithms—PFCM-R, FHAL, and MNFL. While PFCM-R achieved high accuracy with clean data, it struggled with noisy inputs. FHAL showed better noise tolerance but lower precision. PFCM-R is a variant of well-known fuzzy unsupervised learner FCM, and FHAL is the hybrid fuzzy inference engine based on Fuzzy Association Memory and a double-layered FCM we developed. To make the system more robust, we propose the multi-layered neuro-fuzzy learner (MNFL) in this paper, which effectively weakens the association strength between the disease and the observed symptoms, less related to the body part on which the abnormal symptoms are observed. In experiments that are designed to examine how the inference system reacts under increasing noisy input from the user, MNFL achieved 98% accuracy even with non-erroneous inputs, demonstrating superior robustness to other inference engines. This system empowers pet owners to detect health issues early, improving the quality of care and fostering more informed interactions with veterinarians, ultimately enhancing the well-being of companion animals. Full article

Hepatitis C virus (HCV) infection affects over 71 million people worldwide, leading to severe liver diseases, including cirrhosis and hepatocellular carcinoma. The virus’s high mutation rate complicates current antiviral therapies by promoting drug resistance, emphasizing the need for novel therapeutics. Traditional high-throughput screening (HTS) methods are costly, time-consuming, and prone to false positives, underscoring the necessity for more efficient alternatives. Machine learning (ML), particularly quantitative structure–activity relationship (QSAR) modeling, offers a promising solution by predicting compounds’ biological activity based on chemical structures. However, the “black-box” nature of many ML models raises concerns about interpretability, which is critical for understanding drug action mechanisms. To address this, we propose an explainable multi-model stacked classifier (MMSC) for predicting hepatitis C drug candidates. Our approach combines random forests (RF), support vector machines (SVM), gradient boosting machines (GBM), and k-nearest neighbors (KNN) using a logistic regression meta-learner. Trained and tested on a dataset of 495 compounds targeting HCV NS3 protease, the model achieved 94.95% accuracy, 97.40% precision, and a 96.77% F1-score. Using SHAP values, we provided interpretability by identifying key molecular descriptors influencing the model’s predictions. This explainable MMSC approach improves hepatitis C drug discovery, bridging the gap between predictive performance and interpretability while offering actionable insights for researchers. Full article

This paper presents an approach to improving visual object tracking performance by dynamically fusing the results of two trackers, where the scheduling of trackers is determined by a support vector machine (SVM). By classifying the outputs of other trackers, our method learns their behaviors and exploits their complementarity to enhance tracking accuracy and robustness. Our approach consistently surpasses the performance of individual trackers within the ensemble. Despite being trained on only 4 sequences and tested on 144 sequences from the VOTS2023 benchmark, our approach achieves a Q metric of 0.65. Additionally, our fusion strategy demonstrates versatility across different datasets, achieving 73.7 MOTA on MOT17 public detections and 82.8 MOTA on MOT17 private detections. On the MOT20 dataset, it achieves 68.6 MOTA on public detections and 79.7 MOTA on private detections, setting new benchmarks in multi-object tracking. These results highlight the potential of using an ensemble of trackers with a learner-based scheduler to significantly improve tracking performance. Full article

The prediction and analysis of students’ academic performance are essential tools for educators and learners to improve teaching and learning methods. Effective predictive methods assist learners in targeted studying based on forecast results, while effective analytical methods help educators design appropriate educational content. However, in actual educational environments, factors influencing student performance are multidimensional across both temporal and spatial dimensions. Therefore, a student performance prediction and analysis method incorporating multidimensional spatiotemporal features has been proposed in this study. Due to the complexity and nonlinearity of learning behaviors in the educational process, predicting students’ academic performance effectively is challenging. Nevertheless, machine learning algorithms possess significant advantages in handling data complexity and nonlinearity. Initially, a multidimensional spatiotemporal feature dataset was constructed by combining three categories of features: students’ basic information, performance at various stages of the semester, and educational indicators from their places of origin (considering both temporal aspects, i.e., performance at various stages of the semester, and spatial aspects, i.e., educational indicators from their places of origin). Subsequently, six machine learning models were trained using this dataset to predict student performance, and experimental results confirmed their accuracy. Furthermore, SHAP analysis was utilized to extract factors significantly impacting the experimental outcomes. Subsequently, this study conducted data ablation experiments, the results of which proved the rationality of the feature selection in this study. Finally, this study proposed a feasible solution for guiding teaching strategies by integrating spatiotemporal multi-dimensional features in the analysis of student performance prediction in actual teaching processes. Full article

The burgeoning development of generative artificial intelligence (GenAI) has unleashed transformative potential in reshaping English language education. However, the complex interplay of learner, technology, pedagogy, and contextual factors that shape the effectiveness of GenAI-assisted language learning remains underexplored. This study employed a novel mixed-methods approach, integrating qualitative comparative analysis (QCA) and system dynamics (SD) modeling, to unravel the multi-dimensional, dynamic mechanisms underlying the impact of GenAI on English learning outcomes in higher education. Leveraging a sample of 33 English classes at the Harbin Institute of Technology, the QCA results revealed four distinct configurational paths to high and low learning effectiveness, highlighting the necessary and sufficient conditions for optimal GenAI integration. The SD simulation further captured the emergent, nonlinear feedback processes among learner attributes, human–computer interaction, pedagogical practices, and ethical considerations, shedding light on the temporal evolution of the GenAI-empowered language-learning ecosystem. The findings contribute to the theoretical advancement of intelligent language education by constructing an integrative framework encompassing learner, technology, pedagogy, and context dimensions. Practical implications are generated to guide the responsible design, implementation, and optimization of GenAI in English language education, paving the way for learner-centric, adaptive learning experiences in the intelligence era. Full article

Multi-word expressions make up a large proportion of the English language and particularly spoken language. Using multi-word expressions can assist with the impression of fluency, making them useful for language learners to know and use. However, proven methods for teaching this language phenomenon are required, so that learners can easily use multi-word expressions in their conversations. The purpose of the study was to examine the efficacy of a fluency workshop focused on multi-word expression use in conversation and to determine the most appropriate implementation for the Japanese context. An action research structure was used over three iterations of the fluency workshop, learner feedback and teacher observations were used to make improvements. Learner feedback regarding the usefulness of each activity for learning English was compared between the original cohort and subsequent cohorts. The results showed significant differences in levels of perceived usefulness for activities where improvements were made, but also for some activities where no specific improvement was made, suggesting that teaching improves through practice. Pedagogical implications include maximising the time on task via clear instructions, providing visual time constraints, and offering scaffolding to support the use of multi-word expressions when recall seems beyond a learner. Full article

Science, technology, engineering, and math (STEM) education is challenged by industries to incorporate business, engineering, and communication experiences to prepare students for workplace success. In this study, we outline an approach—the STEM Oriented Alliance for Research (SOAR)—to enhance student experience by offering interdisciplinary project-based learning (IPBL) for undergraduate students majoring in electrical engineering, communications, and marketing. We examined how students’ disciplinary and cooperative orientations toward learning shifted in response to their experiences in a semester-long interdisciplinary project-based learning experience with authentic industry outputs. Using a multi-method approach, we explored how interdisciplinary projects influenced student experiences in terms of five collaboration abilities: positive interdependence, accountability, promotive interaction, group processing, and social skills. Further, we observed a shift from fixed- to more growth-oriented mindsets, and from a primarily disciplinary to interdisciplinary focus for their future professional work. The outcomes of the SOAR project make clear that providing structure for professional cooperation on interdisciplinary projects can have profound effects on how students learn to cooperate and position themselves as learners. For most SOAR participants, the experience was deeply formative and contributed to their readiness to cooperate and learn within the interdisciplinary and STEM-oriented workforce. Full article

The number of students enrolling in distance learning programmes is rising worldwide, making distance education (DE) a significant part of higher education (HE). Transitioning into a study programme involves numerous challenges, especially for distance learners who face higher dropout rates and compromised academic performance compared to traditional on-campus students. However, when students master these challenges, study success becomes more likely. Nevertheless, knowledge about transitioning into DE remains limited. This scoping review aims to compile existing knowledge and enhance understanding of the critical initial phase of DE by answering the research question: “What is known about the transition into DE in HE?”. Following the methodological steps outlined in the PRISMA-ScR checklist, we identified 60 sources from five databases, meeting inclusion criteria through a multi-stage screening process. These articles were analysed using qualitative content analysis. We developed a category system with 12 main categories: 1. Process of transition into DE; 2. Reasons for choosing DE; 3. Characteristics of distance learners; 4. Academic success and failure; 5. General assessment of DE; 6. Differences between face-to-face and DE; 7. Advantages of DE; 8. Challenges of DE; 9. Critical life events; 10. Coping strategies; 11. Add-on initiatives; and 12. Recommendations for DE. The results underline the complexity of the transition into DE, which has unique patterns for each student. The article concludes with practical implications and recommendations for supporting the transition into DE. Full article

With the development of financial technology, the traditional experience-based and single-network credit default prediction model can no longer meet the current needs. This manuscript proposes a credit default prediction model based on TabNeT-Stacking. First, use the PyTorch deep learning framework to construct an improved TabNet structure. The multi-population genetic algorithm is used to optimize the Attention Transformer automatic feature selection module. The particle swarm algorithm is used to optimize the hyperparameter selection and achieve automatic parameter search. Finally, Stacking ensemble learning is used, and the improved TabNet is used to extract features. XGBoost (eXtreme Gradient Boosting), LightGBM (Light Gradient Boosting Machine), CatBoost (Category Boosting), KNN (K-NearestNeighbor), and SVM (Support Vector Machine) are selected as the first-layer base learners, and XGBoost is used as the second-layer meta-learner. The experimental results show that compared with original models, the credit default prediction model proposed in this manuscript outperforms the comparison models in terms of accuracy, precision, recall, F1 score, and AUC (Area Under the Curve) of credit default prediction results. Full article