On the Use of Machine Learning and Key Performance Indicators for Urban Planning and Design

Abstract

Global efforts to achieve climate neutrality increasingly rely on innovative urban planning and design strategies. This study focuses on the identification and application of key performance indicators (KPIs) to support policymakers and local authorities in driving sustainable urban transitions. Using a real-life case study of European cities and countries, this research leverages data analytics and machine learning to inform decision-making processes. Specifically, the k-means clustering algorithm was employed to group countries based on socioeconomic and environmental KPIs, while principal component analysis was used to rank the most influential indicators in shaping these clusters. The analysis highlighted GDP per capita, corruption perception, and climate-related expenditure as key drivers of clustering. Additionally, time series analysis of KPI trends demonstrated the impact of policy decisions over time. This study showcases how machine learning and data-driven approaches can provide valuable insights for urban planners, offering a robust framework for evaluating and improving climate-neutrality strategies at both city and country levels.

1. Introduction

With more than half of the global population now residing in urban areas (https://www.worldbank.org/en/topic/urbandevelopment/overview, accessed on 27 September 2024), urban planning and design are fundamental tasks in shaping the quality of life in cities [1,2] and are closely related to sustainability development goals (SDGs). They provide essential infrastructure to promote social unity and environmental sustainability [3,4]. These tasks influence every aspect of urban life and are essential for creating functional and efficient cities. Effective planning ensures that cities have well-designed transportation networks, adequate housing, and accessible amenities such as parks, schools, and healthcare facilities. This not only improves the overall livability of cities but also enhances economic productivity by reducing congestion and improving access to opportunities [2]. Moreover, urban planning and design are critical for promoting social equity and inclusivity [5,6]. By ensuring that neighborhoods are well-connected and provide equal access to services and resources, urban planners can help reduce socioeconomic disparities and create more equitable cities. This includes designing public spaces that are accessible to people of all ages and abilities and fostering mixed-use developments that promote social interaction and diversity [2]. Additionally, these tasks are crucial in addressing environmental challenges such as climate change and pollution [7,8]. Sustainable urban design principles, such as green building practices, efficient public transportation systems, and green spaces, can help reduce greenhouse gas emissions, improve air quality, and enhance urban resilience to climate-related risks. Hence, these tasks are integrated with resilience principles for greater benefits [9,10].

However, adapting urban planning and design strategies poses challenges that require innovative solutions [11]. These challenges include community engagement, information gaps, and uncertainty, among others. Thus, holistic perspectives are required to tackle several urban life aspects in urban planning. For example, a detailed analysis of integrative approaches across different sectors is required for the transition to climate neutrality [12]. These approaches are not limited to targeting renewable energy power but integrate other issues related to transport as well as industry. Decision makers worldwide, including those in advanced nations, are confronted with achieving sustainable environmental configurations, resulting in increased energy consumption, environmental degradation, pollution, inefficient infrastructure and resource management, inadequate decision-making, and socioeconomic inequalities [13]. As a result, several initiatives have been identified to facilitate the integration of urban planning and design solutions [14].

In this context, key performance indicators (KPIs) are defined to evaluate the insight of selected solutions in urban planning and design strategies [15]. KPIs represent data that need to be linked to business strategies and objectives. For instance, Quijano et al. [16] studied smart cities and proposed a KPI-driven evaluation framework. Their approach aims to evaluate the sustainability of cities in a holistic way. Also, Soriano-Gonzalez et al. [17] analyzed KPIs related to transportation and logistics in Barcelona. In their study, they used these KPIs to provide recommendations for policymakers regarding sustainable mobility. Nowadays, the Internet of Things (IoT) replaces the traditional ways of collecting data and storing them [18]. Additionally, the technologies in IoT enable the visualization of the collected data. The data trigger the utilization of advanced techniques for their analysis and visualization, such as artificial intelligence (AI) and, in particular, machine learning (ML). Using data and KPIs, algorithmic approaches have recently been proposed for adaptation in urban planning and design [19,20,21]. Still, the use of ML in urban planning and design is just emerging, and more case studies and examples are required. Given this scenario, the main contributions of this paper are highlighted as follows: (i) based on a real-life and large-scale case study on urban planning and design, we define a series of KPIs to measure the situation of different European cities; (ii) making use of these KPIs, open-access data, and different ML methods, we analyze the situation of each of these cities in comparison to KPI values associated to other cities and countries; and (iii) from the previous analysis, we extract recommendations for policymakers.

The remainder of this paper is structured as follows: Section 2 provides a literature review on how data analytics and machine learning methods have been applied in urban planning and design. Section 3 describes an example of an ongoing project oriented toward urban planning and design, which will serve as a base case study to illustrate our analysis. Section 4 provides a methodology to generate KPIs in the context of the aforementioned case study. In Section 5, open source data are used for the analysis, which results are described in Section 6. Section 7 presents derived managerial insights from the presented case study. Finally, Section 8 summarizes the analysis and highlights the future work.

2. Data Analytics and Machine Learning Methods in Urban Planning and Design

Data analytics refers to examining a large set of data to uncover patterns, correlations, trends, and valuable insights. It involves applying various statistical and computational techniques to understand and interpret data in order to make informed decisions, solve problems, or identify opportunities for improvement [22]. Kandt and Batty [23] identified the characteristics of data analytics in urban planning as follows: (i) interpreting patterns; (ii) deploying computing technologies; (iii) capturing real-time and dynamic changes; and (iv) capturing the compound nature of the data and their mutual impact on each other. Data analytics methods can be employed to understand urban dynamics, assess current status, and predict future trends. Data analytics methods involve descriptive and exploratory analysis, predictive analytics (which includes ML methods), and prescriptive analytics (which includes AI algorithms). Descriptive analytics provides insights into the current state of urban areas [24]. In the context of urban planning, the collected data are KPIs related to area utilization, infrastructure distribution, and environmental factors. Exploratory analytics try to identify patterns within urban data to uncover insights and trends. These techniques are used to analyze the distributions of population, area utilization, and infrastructure. The trends indicate the impact of changes resulting from urban development and provide a guide for planning transportation and infrastructure development. For example, Rybarczyk et al. [25] utilized such analysis in their study to indicate urban design factors influencing cycling.

Predictive analytics involves predicting and forecasting future trends based on historical data using ML and time series, such as the forecasting of generated waste and gas emissions. These predictions enable policymakers to address future resulting challenges and plan proactive actions [26]. Finally, prescriptive analytics recommends actions to optimize urban development and resource allocation, such as optimization models to enhance public transportation routes or waste collection activity. Likewise, simulation models evaluate different scenarios and their impacts, guiding decisions on infrastructure investment and land use planning [27]. Particularly in the area of predictive analytics, ML methods offer powerful tools for analyzing complex urban data and predicting future trends [28]. These techniques are employed to build predictive models for various urban scenarios, such as regression models for predicting gas emissions based on historical data. Additionally, ML models can predict air quality degradation, which helps policymakers develop strategies that mitigate the impacts of climate change. These predictive models provide valuable insights into future urban dynamics, allowing for proactive actions. ML algorithms are capable of recognizing patterns and correlations within data. Thus, indicators could be used to classify urban areas based on their similarities.

Various data analytics and ML techniques could be employed in the context of a large-scale and international project to understand urban dynamics, assess current status, and predict future trends. The integration of ML in the analysis of data speeds up the evaluation process, allowing for more efficient policy recommendations and resource optimization. According to [29], when KPIs are combined with ML, decision-making accuracy and efficiency are greatly enhanced, ensuring actions are aligned with strategic objectives and minimizing biases.

3. Case Study: The UP2030 Project

In order to illustrate with a real-life case study the methodologies and concepts proposed in this paper, an ongoing and large-scale urban planning and design project is introduced. The UP2030 project aims to lead the urban transition to climate neutrality through a strategic urban planning (SUP) approach (https://up2030-he.eu/about-project/, accessed on 27 September 2024). This approach aims to update policies, enhance skills, and prototype solutions for broad implementation. The 5UP concept outlines a strategy involving updating policies (UP-dating), building stakeholder capacities (UP-skilling), prototyping solutions (UP-grading), scaling up impacts (UP-scaling), and fostering communities of practice (UP-taking). Ten pilots are identified in various cities: Budapest, Granollers, Lisbon, Milan, Muenster, Rotterdam, Thessaloniki, Zagreb, Belfast, and Istanbul. A pilot defines a subarea (district) in the involved city.

Reflecting the complexity of urban interactions and the importance of inclusiveness and spatial justice, the project focuses on technological and social innovations to redirect urban planning and design toward sustainability [30]. In this context, KPIs allow for the assessment of adopted policies and the progress toward climate neutrality goals. KPIs could be used to guide the selection of climate neutrality strategies to be adopted by policymakers [31]. Jiang and Kurnitski [32] proposed a framework consisting of 20 KPIs to monitor carbon footprint. The framework was tested on a university campus and was used to propose areas for improvement and accordingly recommendations. In order to enhance policy effectiveness, evidence-based policymaking is crucial, requiring comprehensive evaluation based on data collection and its analysis [33]. The definition of KPIs and the process of data collection pose challenges to the evaluation of urban policies [34]. Additionally, this evaluation requires expertise, which is often limited. These challenges are being addressed through advancements in data analytics and machine learning techniques, which offer tools for data analysis and facilitate decision-making and policy development [15]. Also, the current availability of open data improves access to information, supporting evidence-based policymaking [17].

4. Determination of KPIs

KPIs have a long-standing presence in business management, functioning as essential tools for translating organizational strategy into measurable actions. KPIs serve as guides toward achieving objectives, since they represent the quantifiable aspects of the project’s strategic goals, ensuring alignment with the overarching mission of urban transformation. To achieve operational success, a framework for the definition, monitoring, and evaluation of KPIs is defined as illustrated in Figure 1.

The defined KPIs are classified as core KPIs and pilot-specific KPIs. On the one hand, the core KPIs are derived from project objectives. On the other hand, pilot-specific KPIs are tailored to pilots and reflect pilot’s objectives. The framework in Figure 1 consists of three phases. Phase I initiates the process by drafting a list of KPIs for subsequent phases and involves the following steps:

• Strategic profiling of pilots or projects: Understanding each pilot city or project to define profiles and ensuring KPIs dimension to measure reflect local needs and strategic objectives.
• Validation of pilot profiles: Ensuring defined profiles align with project intent and pilot characteristics through stakeholder consultation, review of strategic documents, profile refinement, and reflecting gained feedback.
• Preliminary KPI formulation: Developing a preliminary list of KPIs categorized into core (common) and pilot-specific KPIs is crucial for monitoring project success.

The profile matches the pilot’s vision and objective on one side and considers dimensions with their categories on the other side. The dimensions are economy, environment, society and culture, governance, and propagation, which are derived from urban planning dimensions. Phase II involves engaging with cities and liaisons to confirm the relevance and practicality of the preliminary set of KPIs. Each KPI is broken down into the required variables to measure it. For example, the percentage of green areas is calculated by dividing the area of green areas by the total city area. These two variables (green area and city area) should be recorded and available for measuring the prescribed KPI. The identification of KPI measurability is fundamental in the KPI identification process. Accordingly, the final list of KPIs is defined. The last phase (Phase III) is the execution phase involving data acquisition and comprehensive data analysis. The framework in Figure 1 is iterative, reflecting the dynamic nature of urban development projects. It adapts, evolves, and responds to changes in project implementation over the three phases of the monitoring process. This iteration is presented by the feedback between the phases. The update involves the defined profile and KPIs. Figure 2 demonstrates the alignment of the KPI monitoring process with the project run and the integration of the process with required inputs, such as defined objectives, storylines, and visions. These inputs facilitate the base for profile definition and the confirmation of the profiles. The KPI sources for this study are standards and technical reports that offer a rich KPI list for urban planning and design, such as the United for Smart Sustainable Cities report (https://unece.org/DAM/hlm/documents/Publications/U4SSC-CollectionMethodologyforKPIfoSSC-2017.pdf, accessed on 27 September 2024) and the World Benchmarking Alliance report (https://assets.worldbenchmarkingalliance.org/app/uploads/2021/07/Just-Transition-Methodology.pdf, accessed on 27 September 2024).

In addition to the direct impact of processes as depicted in Figure 2, some processes influence others. Objectives and visions are formulated during the project. These objectives and visions are the base for defining KPIs as well as for the reporting of KPI analysis. Defined KPIs have to be related to the set objectives and visions, and future KPI values might be compared to the set targets in objective definition during the analysis. The analysis is performed at a later phase and the scope of the analysis targets highlighting the achievements in the context of the defined objectives. The feedback loops in Figure 2 reflect the update of KPIs based on the monitoring outcome.

The integrated list of determined KPIs forms an extensive collection, with some KPIs being unique to specific pilots while others are shared across multiple pilots. The number of pilots sharing common KPIs depends on their objectives. According to the following framework, the KPIs are measurable and relevant to the set objectives. After identifying the KPIs, time frames for tracking and measuring them are established. Once the schedule is set, data collection for the KPI values begins. Further analysis is then performed on the collected data to gain insights, which are valuable for policymakers. This analysis may include predicting future trends based on historical data, benchmarking values across different years or against reference points, or clustering pilots. As a result, proactive and corrective actions can be taken to adjust expected KPI trends, adopt successful strategies from other cities, and assess the progress of pilot projects.

5. Analysis of Open Source Data Availability

This section explores the open data sources of cities and countries involved in our case study. First, the Eurostat dataset (https://ec.europa.eu/eurostat/web/main/home, accessed on 27 September 2024) is examined to look for listed KPIs for the considered cities and countries. Some cities have their own open data portal with some KPIs listed in it. However, these portals do not follow a unified structure, and most of them are written in the country’s mother tongue (with no option to change the language). On the contrary, the Eurostat database provides a source of data that is unified in its representation. The explored KPIs are listed in

Table 1. Considered KPIs.

Abbreviation	KPI	Description
PTN	Public transport network convenience	The length of the public transport network per 100,000 inhabitants
OAM	Opportunity for active mobility	The percentage of the city’s infrastructure dedicated to promoting active mobility
LCE	Low-carbon emission passenger vehicles	The percentage of low-carbon emission passenger vehicles in the city
MWR	Municipal waste rate	The amount of waste generated by households, businesses, and institutions within a municipality
MWRe	Municipal solid waste recycling rate	The percentage of municipal solid waste that is recycled compared to the total waste generated
UER	Unemployment rate	The percentage of the total city labor force that is unemployed
GHGe	GHG emissions	The amount of greenhouse gas emissions produced per capita
REC	Renewable energy consumption	The percentage of renewable energy consumed
NE	Noise exposure	The percentage of city inhabitants exposed to excessive noise levels (greater than 55 dB(A))
GINI	Social polarization/Gini coefficient	The inequality among values of a frequency distribution, representing the extent of income inequality in a population
VCR	Violent crime rate	the violent crime rate per 100,000 inhabitants
CP	Corruption perceptions index	the score for the perceived level of corruption

. In the Eurostat dataset, most KPIs can be retrieved in different units depending on the KPI and specific needs, and some KPIs are derived by combining several recorded values in the dataset—each KPI requires basic variables for its calculation (Section 4). For example, low-carbon emission passenger vehicles include electric, hybrid, fuel cell, and plug-in vehicles. Therefore, the total number of low-carbon emission vehicles is calculated by summing these categories. The dataset is explored to assess the availability of KPIs for the cities.

KPIs in Table 1 are categorized into four dimensions related to urban planning and design aspects:

• Economy dimension: This dimension is divided into three categories: (i) transportation category, which includes public transport network convenience (PTN), the opportunity for active mobility (OAM), and low-carbon emission passenger vehicles (LCE); (ii) waste category, which includes municipal waste rate (MWR) and municipal solid waste recycling rate (MWRe); and (iii) employment category, which includes unemployment rate (UER).
• Environment dimension: This dimension includes GHG emissions (GHGes), renewable energy consumption (REC), and noise exposure (NE).
• Social and culture dimension, which includes the Gini coefficient (GINI) and the violent crime rate (VCR).
• Governance dimension: This dimension is represented by a corruption perception index (CI) in the selected KPIs list.

The availability of recorded KPI values is represented by the number of recorded years and the continuity of the records.

Table 2. Number of KPI-recorded years for each city.

	Belfast	Budapest	Granollers	Istanbul	Lisbon	Milan	Muenster	Rotterdam	Thessaloniki	Zagreb
PTN	–	–	–	–	2	1	8	1	–	–
OAM	–	10	–	–	–	2	6	2	5	10
LCE	–	–	–	–	–	–	–	–	–	–
MDS	–	–	–	–	–	–	–	–	–	–
TM	–	–	–	–	–	–	–	–	–	–
BEE	–	–	–	–	–	–	–	–	–	–
MWR	3	13	–	–	12	3	16	6	–	5
MWRe	–	–	–	–	–	–	–	–	–	–
UER	–	7	8	1	2	5	18	16	3	12
GA	–	–	–	–	–	–	–	–	1	–
PI	–	–	–	–	–	–	–	–	–	–
AQ	–	–	–	–	–	–	–	–	–	–
GHGe	–	12	–	–	–	–	14	25	5	1
REC	–	–	–	–	–	–	–	–	–	–
NE	–	–	–	–	–	–	–	–	–	–
GINI	–	–	–	–	–	–	–	–	–	–
VCR	–	–	–	–	–	–	–	–	–	–
CP	–	–	–	–	–	–	–	–	–	–

and

Table 3. Number of KPI-recorded years for each capital city of the countries in Figure 3.

	London	Budapest	Madrid	Ankara	Lisbon	Roma	Berlin	Amsterdam	Athina	Zagreb
PTN	1	–	3	–	2	2	9	1	–	–
OAM	–	10	3	–	–	9	18	2	5	10
LCE	–	–	–	–	–	–	–	–	–	–
MWR	4	13	3	–	12	13	16	6	–	5
MWRe	–	–	–	–	–	–	–	–	–	–
UER	7	7	24	1	2	7	21	16	–	12
GHGe	25	12	22	–	–	16	25	25	–	1
REC	–	–	–	–	–	–	–	–	–	–
NE	–	–	–	–	–	–	–	–	–	–
GINI	–	–	–	–	–	–	–	–	–	–
VCR	–	–	–	–	–	–	–	–	–	–
CP	–	–	–	–	–	–	–	–	–	–

, as well as Figure 3 record the availability of these KPIs in the Eurostat dataset for cities, their countries, and capitals, respectively. It is seen that not all KPIs are tracked for all countries and cities in the dataset. The majority of the cities have a limited number of records or no records of the corresponding KPIs. In general, the number of recorded years per KPI is greater for countries. Additionally, it has been noticed that KPI values recorded for the United Kingdom were limited after 2018 due to Brexit.

One of the most tracked KPIs is greenhouse gas emissions (GHGes). These records reflect the significant interest and growing awareness in tracking this KPI over several decades. It is monitored at both the country and city levels. Similarly, awareness of the need for waste recycling has increased, leading to the development of several initiatives in this context. This is indicated by the records of MWR and MWRe at the country level (Figure 3). Recorded KPI values on an integrated level (country) represent the average or total recorded values in cities. Examples of these KPIs are those related to the social and cultural dimensions. For the examined cities and capitals, GINI and VCR are not recorded, but they are recorded for most of the countries (Figure 3). The GINI index is recorded for $100\%$ of the countries, and $90\%$ of the countries have documented VCR values. A similar case is noticed for the governance dimension represented by CP. Thus, the individual city’s status is difficult to evaluate with respect to the social polarization given by the GINI value, which is recorded for the whole country. The integrated values mask the differences between big-sized and small-sized cities and the differences between different countries and regions. Notice that the recorded years for these two dimensions are approximately the same for a KPI. Twelve years are recorded for CP, and around nine to ten years for GINI. An exception is found in the United Kingdom instance.

Another possible analysis involves comparing the progress of cities involved in the project with their respective capital and countries. This approach enables the evaluation of data availability and monitoring of a city’s progress in terms of sustainability across the four defined dimensions of urban planning and design. Data collection is typically more consistent at the national level, which poses challenges for cities in gathering and reporting data at a local level. By comparing the recorded KPIs of a city, its capital, and its country, differences in data collection practices and sustainability progress can be identified, offering insights into how local and national efforts align or differ across these dimensions. For example, Figure 4 illustrates a comparative analysis between Milan, Rome, and Italy, using KPIs categorized within the dimensions of transport, waste, employment, environment, social, and organization. The radar plot shows that Milan and Rome exhibit similar values in areas such as transport and employment, while Italy, as a country, has higher values in categories like organization and environment. This analysis highlights the discrepancies in data collection and reporting between the city and national levels, especially regarding sustainability progress across the four dimensions.

6. Empirical Results

To demonstrate the analysis using KPIs, 57 KPIs for European countries for the year 2021 were considered. The KPI list is not limited to those in Table 1 but is supplemented with additional KPIs, some of which are shown in

Table 4. Selected KPIs.

Code	KPI	Unit
CC	Contribution to climate-related spending	Million euro
CC_E	Control of corruption: estimate	Number
EE	Unemployment expenditure	Euro/inhabitant
VA_E	Voice and accountability, freedom of expression: estimate	Number
GDP_PC	GDP/capita	USD
EH	Expenditure on Sickness/healthcare and disability	Euro/inhabitant
PCECB	Population with confidence in EU institutions	Percentage
GE_PR	Governance effectiveness: percentile rank	Percentage
RC	Recycling of batteries and accumulators, cadmium content	Tonne
EHS	Housing and social exclusion expenditure	Euro/inhabitant
ET1	Final energy consumption in transport by type of fuel (gas and diesel oil)	Number
IPS	E-government activities of individuals via websites	Percentage
RL_E	Rule of low: estimate	Number
zeroEV	Total number of electricity, hydrogen, and fuel cells passenger cars	Number

.

Different types of analysis could be applied to KPIs. Soriano-Gonzalez et al. [17] demonstrated the use of such techniques to support policymakers. In this section, clustering techniques are used. This clustering enables policymakers and other stakeholders to make more informed and targeted decisions, optimizing resources and efforts to achieve a positive and sustainable impact. Machine learning offers a variety of clustering algorithms, such as k-means, hierarchical clustering, and DBSCAN. Clustering algorithms assign observations to clusters while minimizing the Euclidean distance between the cluster’s observations and the centroid. The k-means clustering algorithm starts with the random selection of k centroids, representing the centers of the initial clusters [35]. Then, it reassigns each observation to the cluster with the closest centroid during each iteration. Subsequently, the centroids are recalculated as the mean of all observations within each cluster. This iterative process continues until the centroids converge, i.e., when they do not show significant changes in successive iterations, thus ensuring the optimality of the data partitioning [36]. In addition to k-means, hierarchical clustering and density-based spatial clustering of applications with noise (DBSCAN) are widely used clustering methods. On the one hand, hierarchical clustering builds a hierarchy of clusters either through an agglomerative (bottom-up) or divisive (top-down) approach, which is particularly useful when the number of clusters is not known a priori [37]. On the other hand, DBSCAN identifies clusters based on areas of high density and is particularly effective in discovering clusters of arbitrary shapes while also handling noise in the data [38]. All clustering experiments were conducted using Python version $3.11$. The implementation of k-means, hierarchical clustering, and DBSCAN was carried out with standard libraries available in Python’s machine learning ecosystem. After applying the three algorithms to the dataset, it was observed that the clustering results were identical across all methods, with each algorithm returning the same cluster assignments. Given this consistency in the results, it was decided to present only the k-means results in this paper, as they are representative of the overall clustering performance.

This study applies the aforementioned clustering algorithm to a dataset that includes socioeconomic and environmental indicators from several European countries. This approach allows the identification of homogeneous clusters, facilitating the analysis of patterns and trends within each group. Incremental principal component analysis (PCA) was applied to the data to reduce the data dimensionality. PCA helps to identify the main influencing factors by extracting relevant characteristics and eliminating redundant parameters that do not contribute significantly to decision-making [39]. This technique improves the analysis by focusing on the variables with the greatest impact. This study revealed that most of the variability in the data was captured in the first four principal components. For visualization purposes, the dimensionality was reduced to two components. This reduction helps to represent the clusters in a two-dimensional space clearly. To carry out the analysis of which KPIs have most influenced the construction of the clusters, the importance of each of the 57 KPIs was evaluated, identifying the ten most relevant. Figure 5 shows the most important KPIs and their relative percentage of importance. Some of the KPIs that stand out are GDP per capita (GDP_PC), corruption perception index (CP), and contribution to the international commitment of USD 100 billion in climate-related expenditures (CC), which have the highest impact on cluster formation compared to the other KPIs.

Figure 6 shows the distribution of the countries within each cluster. The bar chart on the left illustrates the number of countries in each cluster, while the pie chart on the right displays the percentage distribution. Cluster 0 contains 14 countries, representing $46.9\%$ of the total. Cluster 1 comprises 10 countries, accounting for $34.4\%$. Cluster 2 has 4 countries, making up $12.5\%$, and Cluster 3 consists of 2 countries, representing $6.2\%$ of the total.

Figure 7 shows the projection of the countries in a two-dimensional space using PCA. This representation allows visualizing the dispersion and grouping of the countries according to the identified clusters. The axes of the graph represent the first two principal components, which explain most of the variability in the data. Cluster 0 includes 15 countries mainly from Central and Eastern Europe, such as Bulgaria, Czech Republic, Greece, Croatia, Cyprus, Latvia, Lithuania, Hungary, Malta, Portugal, Romania, Slovenia, Slovakia, Albania and Serbia. These countries present characteristics of moderate development. GDP per capita values vary, with urbanization levels reflecting progress toward modernization. Indicators of access to drinking water and sanitation are also diverse, with some countries reaching almost full coverage, while others are still developing. Expected years of education and governance levels show moderate values, suggesting areas for improvement in infrastructure and public services.

Cluster 1 groups 11 mostly Northern European countries, including Belgium, Denmark, Estonia, Ireland, Luxembourg, the Netherlands, Austria, Finland, Sweden, Iceland and Norway. These countries stand out for having the highest GDP per capita and the best rates of government effectiveness. They have high levels of urbanization, almost universal access to water and sanitation, and the highest expected years of education. These indicators reflect advanced economies with solid infrastructure and high living standards. Cluster 2 consists of 4 countries: Spain, Italy, Poland and Turkey. These countries have the second-highest levels of GDP per capita and expected education. Although largely urbanized, they show variations in other social and economic indicators. Access to drinking water and sanitation is high but does not reach the near-universal levels observed in Cluster 1. This suggests that, while these countries have developed infrastructure, there are areas that could benefit from further improvement. Cluster 3 includes Germany and France, two of the largest economies in Europe. These countries exhibit high GDP per capita, high levels of urbanization, and near-universal access to water and sanitation. They also have strong governance indices and high expected years of education, indicating well-established socioeconomic conditions and robust infrastructure. These countries lead in several aspects of development, positioning them as role models for other European nations.

7. Managerial Insights

The defined clusters of the countries using the ML technique match the perspective of the countries’ differences and similarities. The KPI values at the specified time, the year 2021, are used as the base for the definition of the clusters. These clusters represent the result of different decisions made in the context of urban planning and design in the last decades. Realized decisions are translated into KPI value changes in the subsequent years. Thus, the KPI values for months from January 2008 to May 2024 (https://ec.europa.eu/eurostat/web/main/data/database, accessed on 27 September 2024) are tracked and demonstrated in Figure 8 and Figure 9. The time series analysis of various KPIs reveals significant trends and patterns. The electricity consumption trends depicted in Figure 8a show that Germany maintains significantly higher levels compared to Greece, Belgium, and Spain. This disparity reflects differences in economic scale and the extent of electrification and industrialization. In Figure 8b, a comparison of electricity consumption within Cluster 2 countries (Turkey, Poland, Spain, and Italy) reveals that Italy and Spain not only consume more electricity but also exhibit greater fluctuations, pointing to a more dynamic industrial sector and varying electricity demand throughout the year. The consistently higher consumption levels in Italy and Spain emphasize the importance of investing in renewable energy sources and improving energy efficiency to meet sustainability goals and reduce dependency on fossil fuels.

Unemployment trends, as shown in Figure 9a, highlight the socioeconomic challenges faced by different clusters. Spain and Greece, both severely affected by the 2008 financial crisis, experienced high unemployment rates that gradually decreased over the years. Belgium’s more stable and lower unemployment rates contrast sharply, indicating more resilient labor markets. In Figure 9b, the comparison within Cluster 2 countries reveals that Spain consistently faced the highest unemployment challenges compared to Turkey, Poland, and Italy, reflecting structural issues within its labor market. These insights are crucial for formulating targeted employment policies and economic reforms to enhance job creation and labor market stability.

These insights have significant implications for policymakers and urban planners. Understanding the clustering and associated KPIs trends allows for an effective allocation of resources and targeted interventions. For instance, countries in Cluster 0, which face moderate development challenges, can benefit from policies focused on enhancing infrastructure and economic diversification, such as those that have been implemented by countries in Cluster 3. In contrast, countries in Cluster 3, like Germany and France, can lead initiatives in innovation and sustainable practices, serving as models for other nations. The alignment of KPI trends with the Agenda 2030 goals highlights the critical areas where progress is needed to achieve sustainable development. For example, the emphasis on reducing unemployment aligns with SDG 8 (decent work and economic growth), while managing energy consumption and promoting renewable energy sources support SDG 7 (affordable and clean energy). The data-driven approach using clustering techniques provides a robust framework for monitoring and evaluating progress toward these goals. Policymakers have the opportunity to design strategies that are specifically tailored to the unique needs and characteristics of each cluster. This could involve creating economic development policies that align more closely with the socioeconomic and environmental profiles of different country groups [40]. Such targeted approaches ensure that the policies are not only relevant but also impactful. Furthermore, government agencies and international organizations can enhance their efficiency by directing resources—both funds and technical support—precisely to the countries within each cluster that need them the most. This method of allocation maximizes the effectiveness of development aid and support programs. Sharing best practices and establishing collaborations become easier for countries within the same cluster [41]. Joint projects in technology, environment, infrastructure, and other key sectors can drive sustainable development. These collaborations foster innovations and improvements that benefit all countries involved [42].

Likewise, non-governmental organizations and research institutions can strategically plan their interventions and projects based on clustering information. By aligning their efforts with the specific needs and realities of each cluster, these organizations can achieve greater effectiveness and sustainability in their initiatives. The use of clusters also provides a solid foundation for monitoring and evaluating policies and programs over time [43]. This enables stakeholders to measure the impact of their interventions and adjust their strategies accordingly based on observed results. Continuous monitoring ensures that policies remain effective and relevant. Moreover, businesses and investors can identify market opportunities in countries that share similar characteristics. Financial institutions and insurers can utilize the clustering insights to better assess and manage risks. Understanding the socioeconomic and environmental dynamics of each group of countries enables more informed decisions about lending, investment, and insurance. This leads to more stable financial systems and better support for sustainable development. Additionally, clustering analysis can highlight critical areas requiring urgent intervention, such as regions with high poverty, poor education quality, or healthcare deficiencies. Policymakers can then focus their efforts on these areas to improve living conditions and promote sustainable development, ensuring no region is left behind in the pursuit of the Agenda 2030 goals [44]. This integrated approach to using clustering insights underscores the potential for achieving the United Nations’ SDGs.

8. Conclusions

This study examined the use of KPIs in urban planning and policymaking, focusing on a real-life case study involving European cities and countries. The primary objective was to illustrate how KPIs can support policymakers in tracking changes related to urban planning, governance, and sustainability. Key observations from exploring open-source datasets include the following: (i) KPIs are primarily tracked at the country level rather than at the city level; (ii) global trends and growing awareness of sustainability issues have increased interest in tracking KPIs, although many remain underutilized; (iii) governance, social, and cultural dimensions are typically reported at the national level; and (iv) the historical data available for city-specific KPIs are limited.

This study also employed machine learning techniques to analyze the recorded KPIs, utilizing clustering and visualization methods to provide a clearer understanding of patterns and trends. Through clustering, three key KPIs emerged as the most influential: GDP per capita, the corruption perception index, and contributions to climate-related expenditures. These indicators revealed significant disparities in economic development, access to services, urbanization, and governance effectiveness across European countries. This study’s results align with the Agenda 2030 sustainable development goals, showing that data-driven approaches can inform more effective decision-making.

The time series analysis of KPIs illustrated how the evolution of KPI trends reflects past decisions made by policymakers. This shows that adapting policies can directly affect the future KPI trends, underscoring the importance of continuous monitoring for effective governance. The comparative analysis between city-specific and country-level KPIs highlighted the need for more granular, city-based data, as cities function as distinct governmental units in many cases. Without city-level KPIs, policymakers lack the precise feedback needed to make effective urban planning decisions.

One of this study’s main challenges was the lack of standardized, high-quality datasets for city-level KPIs. Future work will focus on expanding the analysis to include city-specific KPIs, as well as integrating new data sources to improve the completeness and quality of the available information. Additionally, future research will explore the impact of specific urban planning actions on KPIs, using comparative analysis to evaluate the effectiveness of these actions over time.