Actionable insights for equitable urban sustainability transition through building-level assessment of sustainable development goals

Abstract

Traditional Sustainable Development Goals (SDGs) evaluations, based on administrative units, lack the granularity required for effective urban interventions. This research emphasizes equity and sustainability, addressing how concepts like the 15-minute city often overlook accessibility barriers for all residents. A data-driven approach was used to generate building-level sustainability scores (BLSS) in Hong Kong, integrating equity within the 15-minute city framework. A novel simulation framework analyzes the impact of infrastructure development scenarios on BLSS across SDG indicators. Geospatial analysis calculates detailed sustainability scores, while statistical measures like Gini coefficients identify inequities and guide infrastructure improvements. Machine learning simulations explore future urban development scenarios, enabling proactive planning. BLSS, derived from a dataset covering over 40,000 residential buildings and 100 government departments, is scaled and aggregated to compare SDG indices at neighborhood, district, regional, and city-wide levels. This framework equips urban planners and policymakers with a robust tool for data-driven decision-making, fostering equitable and sustainable urban growth.

Introduction

Cities like Copenhagen, Melbourne, Barcelona, and Paris have made notable strides toward sustainable urbanization, each offering a model for creating more accessible, green, and equitable environments through specific policy initiatives. Copenhagen has significantly reduced emissions and traffic by adopting its Carbon Neutrality Plan and building an extensive bike infrastructure¹. However, its reliance on costly investments poses challenges for less affluent cities². Melbourne has transformed neighborhoods like Docklands by implementing the Green Star Communities framework, focusing on sustainable housing and accessibility, but rising property prices threaten housing affordability and equitable access^3,4. Barcelona has introduced its innovative Superblock Program to reduce traffic and pollution by restricting cars in pedestrian-friendly zones, achieved through urban design and public engagement⁵. However, some residents criticize it for disrupting businesses and commuting⁶. Paris has embraced the 15-Minute City model, improving local infrastructure and services to ensure essential amenities are accessible within a 15-min walk or bike ride from every home, reducing the need for long commutes and fostering local economies⁷. Despite these successes, certain neighborhoods in Paris, like the Le Marais, Belleville, and the Canal Saint-Martin corridor, that have experienced substantial investment and renewal under the 15-Minute City model have also seen rising gentrification pressures and elevated living costs, potentially excluding lower-income residents from fully benefiting⁸. While these cities have implemented impressive sustainability measures through their respective policies, they also reveal the social and economic challenges of transitioning to equitable, sustainable urban environments, highlighting the need for broad and inclusive access to critical infrastructures such as affordable housing, public transportation, healthcare, education, and green spaces^9,10.

The concept of a 15-minute city is an urban planning solution that attempts to promote accessibility to essential services and amenities within a short walking or biking distance¹¹. However, it is important to note that the concept of 15-minute cities is not universally implemented, and even where it is, it may not result in equitable access to essential services and amenities. While conceptually, the 15-minute city provides a good framework for urban design, there is a lack of rigorous quantitative methodology that can benchmark accessibility profiles for all residents and identify any hotspots. Environmental justice principles highlight the importance of distributing resources and opportunities fairly, particularly in historically underserved areas¹². Therefore, equitable and prescriptive urban design is needed to ensure all residents live in a 15-minute city and not just a handful of privileged city residents in the center of the city. This paper addresses this limitation in urban planning by prioritizing both equity and sustainability to realize the full potential of the 15-minute city concept. Within the growing discussion on the 15-minute city as a model for sustainable urban development, our prior research has addressed the limitations of the 15-minute city and proposed a framework to promote equity in urban accessibility¹³. Through the utilization of quantitative methodologies such as the Gini coefficient, our studies have systematically examined spatial and socio-economic disparities in residents’ access to essential infrastructures. Our research found the highest Gini coefficients in waste management and elderly care, signaling major access inequities. Hospitals had a lower Gini, indicating better access. These findings highlight the need for targeted interventions, particularly in waste management and elderly care.

Sustainability assessment in cities encompasses diverse methods, including ecological footprint analysis, urban metabolism, life cycle assessment, and multi-criteria decision analysis, among others, each evaluating environmental, social, and economic aspects to gauge a city’s sustainability performance. Multiple frameworks, including the Sustainable Development Goals (SDGs), have been utilized for systematic sustainability assessment. The SDGs provide a comprehensive framework aimed at addressing global challenges, such as poverty, inequality, and climate change¹⁴. The framework endeavors to establish a universal language for sustainable development by offering a structured approach to measuring progress, facilitating benchmarking among nations, and enabling quantitative evaluation of efforts¹⁵. However, despite its overarching scope, the current SDG framework often lacks actionable guidance¹⁶. While the SDGs provide overarching objectives, they may not always offer clear and practical recommendations tailored to the unique contexts and challenges faced by individual cities. This absence of detailed guidance can make it challenging for local governments and stakeholders to translate the broad aspirations of the SDGs into concrete actions and policies at the urban level. This is primarily because SDGs are typically applied at the administrative unit level, which can limit their effectiveness in driving sustainable practices within individual buildings¹⁷. Applying SDG indicators at the building scale allows for more granular measurement and monitoring of progress, enabling stakeholders to better understand the impact of their actions and simulate scenarios for future development¹⁸. Despite increasing recognition of the importance of integrating SDGs at the building level, critical research gaps persist^19,20,21. Challenges include the lack of a comprehensive scoring system to evaluate SDG performance, limited understanding of synergies and trade-offs between SDGs in buildings, and sparse empirical studies on the outcomes of incorporating SDG principles^22,23.

Past studies highlight the potential of building-level data when coupled with advanced analytics and robust statistical techniques^24,25. Several initiatives employ sophisticated tools and technologies, such as Geographic Information Systems (GIS), energy simulation software, remote sensing technology, and data-driven analysis platforms, to conduct detailed assessments of neighborhood sustainability. Initiatives like Global ESG Benchmark for Real Assets (GRESB) assess sustainability across real estate portfolios, providing standardized data for Environmental, Social and Governance (ESG) performance evaluation^26,27,28. Leadership in Energy and Environmental Design (LEED), under the U.S. Green Building Council (USGBC), evaluates neighborhoods comprehensively, considering factors such as site selection, transportation options, and community engagement to foster sustainable, vibrant communities. Similarly, the Hong Kong Green Building Council Limited (HKGBC) assesses neighborhood environmental performance in Hong Kong, emphasizing green spaces, walkability, public transit access, and community involvement to enhance urban sustainability and resident well-being^29,30. In their attempt to assess sustainability, these initiatives tend to focus on individual buildings or portfolios, limiting their impact on broader urban contexts. These initiatives primarily emphasize inward-looking, building-specific parameters such as energy efficiency, thermal shading, lighting, and material performance, often overlooking macro-scale dimensions, such as street connectivity, public infrastructure, and land-use integration, that shape the spatial morphology, functional continuity, and social accessibility of the wider urban environment. While their assessments excel at evaluating individual building performance, they fall short in addressing how these buildings integrate into the wider urban context, limiting their effectiveness in fostering holistic urban sustainability. These initiatives often issue certifications based on specific criteria, but they lack the ability to compare buildings meaningfully, even when they share the same certification. The underlying factors contributing to these certifications are not always transparent or comparable across buildings, making it difficult to determine how different sustainability elements impact overall performance^31,32,33. These frameworks, while valuable for standardizing data and promoting sustainability within specific neighborhoods, fall short in integrating the complex interactions across different scales of urban systems. In contrast, our approach goes beyond isolated certifications by integrating building-level data into a framework that captures the interactions between various sustainability factors. This allows for a more nuanced comparison across buildings and a clearer understanding of how different elements contribute to urban sustainability, thus enabling more effective and comprehensive planning. Our approach leverages building-level data within a more holistic framework that considers both micro- and macro-level factors, allowing for more comprehensive urban sustainability planning that can effectively bridge the gap between localized assessments and broader urban contexts. Thus, there is a need for broader integration of building-level data across urban planning and development initiatives to achieve more comprehensive and impactful urban sustainability goals. Therefore, building-level data emerges as a potent tool for achieving holistic urban sustainability, backed by robust analytics and statistical evidence^34,35,36,37.

Recent scholarly discourse has recognized the potential of merging geospatial modeling with machine learning techniques, both Supervised and Unsupervised algorithms, to address these limitations³⁸. Supervised learning algorithms such as Decision Trees, Support Vector Machine (SVM), K-Nearest Neighbors (KNN), and Perceptron are employed for classification and prediction tasks in geospatial applications, particularly in urban development. These algorithms help predict and classify factors like infrastructure growth, land-use planning, and risks related to urban expansion, including traffic congestion and road safety, as well as health and landslide susceptibility. On the other hand, unsupervised algorithms like k-means clustering are used to identify patterns and group regions in unlabeled geospatial data, especially for environmental risk assessments, where they reveal patterns related to flood risks, landslide susceptibility, soil erosion, and pollutant distribution. Researchers are increasingly acknowledging the significance of moving beyond the conventional neighborhood-level evaluations and delving into finer scales, such as individual buildings, to capture the intricate interplay of diverse factors impacting sustainability^39,40. In this context, the presented research introduces an algorithm that leverages this synergy. By integrating geospatial models and machine learning, the algorithm facilitates the computation of Building-level Sustainability Scores (BLSS). This promising approach holds the capacity to fill critical gaps in current literature, fostering equity in accessibility and promoting justice by providing insights into resource distribution and opportunities within urban areas.

Diverse urban settings, both high-density and lower-density cities, present unique challenges and opportunities to quantify and assess sustainability. According to Cohen, “Sustainability and equity considerations require that a home for a single person should be no larger than 215 square feet”⁴¹. Hong Kong, with its remarkably compact living arrangements averaging just 172 square feet per person, emerges as a prime exemplar of sustainable urban development^42,43. Despite Hong Kong’s current need to expand its compact housing, its model remains a powerful example for future cities aspiring to sustainable living. In high-density cities like Hong Kong, efficient public transportation, well-planned urban layouts, and accessible essential services are crucial for sustainability, minimizing transportation emissions and resource usage^44,45,46. These cities also foster vibrant social interactions, cultural exchange, and economic growth. Conversely, lower-density cities may offer more green space and less congestion but face challenges such as urban sprawl, longer commutes, and higher per capita energy consumption^47,48,49. Achieving sustainable urban environments, regardless of density, hinges on balancing population density with robust infrastructure, effective land-use planning, and resource management strategies^50,51.

Results

Following the discussion of sustainability challenges and strategies in urban environments, the output of this study offers a detailed examination of sustainability metrics at multiple granular levels, providing valuable insights for stakeholders in urban development.

Granular insights: building-level analysis of sustainability in urban contexts

The output of this study can be studied at the smallest granular level, building level. Figure 1 presents a detailed building-level analysis, offering a focused examination of individual structures. The figure provides specific metrics, including individual SDG scores, overall SDG performance, facilities assessment, and building census statistics. This granular evaluation allows for a comprehensive understanding of the sustainability characteristics inherent to each building. By synthesizing diverse data points at the building level, the figure contributes to a more nuanced exploration of the factors influencing sustainability within the urban context. This analytical depth aids in unraveling the intricacies of individual buildings’ contributions to broader sustainability goals, presenting valuable insights for researchers, policymakers, and stakeholders in the realm of urban development.

Fig. 1: Building-level analysis offering insights into individual building metrics, including specific SDG scores, overall SDG performance, facilities assessment, and building census statistics.

This figure provides a comprehensive analysis of the Sang On Building located at 42A Bulkeley Street, Hung Hom. The evaluation includes demographic data, proximity to facilities, and Sustainable Development Goal (SDG) performance. Key elements of the figure are as follows: Building Overview: Address: 42A Bulkeley Street, Hung Hom, Coordinates: 22.3077901, 114.1839482, Average Population: 173 (87 males, 83 females), Average Income: HK$ 14,820, Gender distribution: approximately equal., Average SDG Score: The building has an average SDG score of 3.8, significantly higher than the citywide average of 1.58. Notable contributors to the score include: SDG 1 (No Poverty): 7.96, SDG 10 (Reduced Inequalities): 4.86, SDG 11 (Sustainable Cities): 6.29, SDG 16 (Peace, Justice): 9.97, Facilities Assessment: The radar-style scoring highlights accessibility to essential facilities near the building: Elderly Care: Perfect score of 10, Police: 4.3, Child Care: 3.93, Hospital: 5.96, Waste Management: 3.32, Schools: 2.72, Map Overview: A map provides a spatial representation of key facilities in the neighborhood, including police stations, childcare centers, hospitals, and schools, Icons on the map are used to denote facility types, while the scale bar at the bottom ensures spatial accuracy. Data Table: A right-hand table lists individual SDG scores for the building (e.g., SDG 1 = 7.96, SDG 16 = 9.97) alongside the average city scores for comparison. This figure integrates demographic, economic, and SDG-related metrics to illustrate the building’s unique performance, providing a holistic understanding of its sustainability and access to services in an urban context.

In Fig. 1, the case study of the Sang On building demonstrates a strong overall SDG score of 3.8, significantly above the citywide average of 1.58, on a 0–10 scale where 0 is poor and 10 is excellent. To ensure consistency and comparability, min-max scaling was applied twice during the analysis: first to normalize individual SDG scores, ensuring consistent comparisons across different goals, and then again to calculate the overall SDG score. This method was chosen for its ability to preserve relative distances while standardizing scores, allowing for a transparent and fair evaluation of both individual and combined SDG performance. The notable contributors to this heightened score are identified as SDG 1 (7.96), SDG 10 (4.86), SDG 11 (6.29), and SDG 16 (9.97). Of particular interest are the facility-specific scores, with the highest observed in elderly care at 10 and the lowest for schools at 2.72, delineating the building’s distinctive characteristics in these domains. Demographically, the male-to-female ratio within the building stands roughly at 51:49, suggesting a balanced gender distribution among residents. Furthermore, the median household income within the building is HK$23,000 (approximately US$2960), while the average individual monthly salary is HK$14,820 (approximately US$1900), both aligning closely with citywide income benchmarks. The distinction reflects the difference between total household earnings (median income) and the per capita wage level (average salary), offering a more granular understanding of the socio-economic characteristics of the residents. These nuanced findings offer a refined understanding of an individual building’s unique sustainability profile, contributing valuable insights to the broader discourse on urban development and providing a data-driven basis for informed decision-making and strategic interventions.

Multilevel assessment: comparative analysis of SDG scores across buildings, districts, and regions

The framework discussed in this article provides a detailed analysis and scores at various levels. Figure 2 presents a meticulous multilevel analysis of SDG scores, examining sustainability metrics at the levels of individual buildings, districts, and regions. This granular assessment facilitates the hierarchical ranking of individual buildings based on their SDG scores, allowing for intra-regional and inter-regional comparative evaluations.

Fig. 2: Multilevel analysis of SDG scores—assessing sustainability at the building, district, and regional levels.

This figure showcases a comparative multilevel analysis of Sustainable Development Goal (SDG) scores across buildings, districts, and regions in Kowloon, Hong Kong. The focus is on evaluating the relative sustainability performance of each level, with a particular emphasis on the Sang On Building, Kowloon City District, and surrounding areas. Regional and District Scores: The Kowloon City District achieves the highest overall SDG score of 2.04, signifying strong sustainability efforts and access to essential services. The Kowloon Region demonstrates an overall SDG score of 1.62, highlighting its average performance compared to other regions. The Sang On Building, specifically, stands out with an overall SDG score of 3.8, significantly outperforming both district and regional averages. Map Representation: The map uses a gradient color scheme to represent SDG performance, with darker shades indicating higher scores. The boundary of the Kowloon City District is clearly marked, emphasizing its spatial relationship to the Sang On Building and surrounding infrastructure. Icons and arrows point to key areas of interest, including the highlighted Sang On Building. District-Wide Data Table: The accompanying table provides a detailed breakdown of SDG scores (SDG 1–17) for all districts. Notable observations from the table include: Islands District: Lowest overall SDG score of 1.17, indicating a need for targeted development initiatives. Kowloon City District: Strong scores across SDG 9 (Industry, Innovation, and Infrastructure: 10) and SDG 11 (Sustainable Cities: 4.32). Insights for Policymakers: The figure highlights disparities in SDG performance, offering a data-driven basis for prioritizing development efforts. Districts and regions with lower SDG scores, such as Islands, Eastern, and Sha Tin, present opportunities for targeted improvements. This figure provides a multilevel visualization of urban sustainability, enabling comparisons across buildings, districts, and regions. It supports urban planners and policymakers in designing strategies to enhance SDG performance while addressing inequalities in resource distribution and service accessibility.

The findings reveal that the Kowloon City district attains the highest overall SDG score at 2.04, indicative of commendable sustainability achievements within this locale. Conversely, the Islands district displays the lowest overall SDG score at 1.17, suggesting potential focal points for targeted improvement. In Kowloon, districts including Islands, Eastern, Central, and Western, and Sha Tin are identified as having lower SDG scores compared to others, with an overall score below 1.5. For detailed scores of all the buildings and regions in Hong Kong, refer to Tables 2, 3, and 4 in the supplementary information. This highlights substantial opportunities for improvement within these regions. Thus, policymakers and urban planners can strategically prioritize interventions in these areas to enhance their SDG performance, facilitating more sustainable and equitable urban development.

Regional disparities in sustainability: analyzing top SDG scores across Hong Kong

A drastic score variation is observed throughout Hong Kong City. Figure 3 presents the top SDG scores across various regions in Hong Kong, accompanied by a detailed breakdown of individual SDG contributions.

Fig. 3: Illustration of peak SDG scores in different regions in Hong Kong, highlighting individual SDG contributions to the elevated overall SDG score.

This figure illustrates regional disparities in Sustainable Development Goal (SDG) performance across six regions in Hong Kong, highlighting spatial variations in sustainability metrics and the contributions of individual SDGs to overall scores. Regions and SDG Performance: Region 1 (Hung Hom and Prince Edward): Strong scores in SDG 1 (5.8), SDG 11 (8.2), and SDG 16 (9.8), indicating efforts in poverty alleviation, sustainable cities, and promoting peaceful, inclusive societies. Region 2 (Tai Po and Hung Shui Kiu North): SDG 16 (9.7) and SDG 10 (4.6) show the region’s focus on reducing inequalities and enhancing societal justice. Region 3 (Lam Tsuen, Ping Shan, and Wang Chau): Balanced performance across SDG 4 (Education: 4.8), SDG 10 (6.2), and SDG 16 (7.2). Region 4 (Aberdeen and South Horizons): Moderate scores in SDG 10 (4.3) and SDG 16 (9.5), with room for improvement in other goals. Region 5 (Ma On Shan): Strong in SDG 11 (8.2) and SDG 16 (9), reflecting a focus on sustainable cities and inclusive societies. Region 6 (Sheung Shui): Outstanding performance in SDG 6 (Water and Sanitation), SDG 13 (Climate Action), and SDG 17 (Partnerships), each scoring 10. Spatial Analysis and Trends: The color gradient on the map (0–10 scale) highlights the SDG score intensity, with darker areas indicating higher performance. The map reveals a stark contrast in sustainability achievements across regions, with Kowloon Region displaying the highest average SDG score (1.62), followed by the New Territories (1.47), and Hong Kong Island scoring the lowest (1.38). Dominant SDGs: SDG 16 (Peace, Justice, and Strong Institutions): Consistently high across all regions, with an average score of 9.6, emphasizing societal inclusivity and justice. SDG 1 (No Poverty): Average score of 4.45 across regions, demonstrating significant focus on poverty alleviation. SDG 10 (Reduced Inequalities): An average score of 4.21 highlights ongoing efforts to address disparities. This figure underscores the spatial heterogeneity in SDG attainment across Hong Kong and identifies regions requiring targeted interventions. Policymakers can leverage these insights to address regional imbalances, prioritize SDGs with lower scores, and strengthen sustainability in underperforming areas.

Within most of these regions, SDG 16 consistently emerges as a prominent contributor, with an average score of 9.6 across the entire Hong Kong. This signifies a notable emphasis on promoting peaceful, just, and inclusive societies within the assessed areas. Additionally, SDG 1, addressing poverty eradication, plays a pivotal role in elevating the overall BLSS across several regions. The average SDG 1 score for Hong Kong stands at 4.45, highlighting its significance in the sustainability landscape. Furthermore, the impact of SDG 10, focused on reducing inequality, is discernible, as reflected in the average score of 4.21 across the assessed regions. These observations illuminate key thematic areas that contribute significantly to the sustainable development profile of Hong Kong.

The outcomes derived from the implemented methodology reveal the spatial heterogeneity in SDG scores across distinct regions within Hong Kong. Hong Kong Island exhibits the lowest average SDG score, quantified at 1.38, followed by the New Territories with a marginally higher mean score of 1.47. In contrast, the Kowloon region attains the highest observed average SDG score, registering at 1.62. The SDG scores, measured on a scale of 0 to 10, highlight an overarching lower level of sustainable development attainment within the broader context of Hong Kong.

Age-demographic sustainability hotspots: spatial insights across Hong Kong regions

Figure 4 depicts sustainability hotspots across Hong Kong regions, categorized by age groups.

Fig. 4: Spatial analysis of age-group specific sustainability hotspots in Hong Kong regions.

This figure presents a spatial analysis of Sustainable Development Goal (SDG) performance hotspots across Hong Kong, categorized by age demographics (<15, 15–39, 40–64, and >65) within the regions of Kowloon, Hong Kong Island, and the New Territories. Age-Based Distribution of SDG Scores: <15 Age Group: Higher SDG scores are concentrated in urbanized areas of Kowloon and Hong Kong Island, as seen in the dense clusters of dark red hotspots. The New Territories display fewer high-score areas, suggesting limited sustainability benefits for younger populations. 15–39 Age Group: This age group benefits from the most widespread and evenly distributed hotspots, particularly in Kowloon, which appears to offer the best overall SDG outcomes for this demographic. Sparse high-score regions in the New Territories highlight disparities in sustainable living conditions for younger adults. 40–64 Age Group: A marked decline in high-SDG-score areas is visible across all regions, with notable gaps in Hong Kong Island and the New Territories. The findings suggest older working-age adults may face challenges in accessing equitable and sustainable resources. >65 Age Group: High-SDG hotspots become scarcer, especially in New Territories, indicating reduced access to facilities or infrastructure for elderly residents. Limited density of red areas across all regions implies disparities in sustainable development benefits for the elderly population. Regional Patterns: Kowloon: Demonstrates the most consistent SDG performance across all age groups, with concentrated high-score zones. Hong Kong Island: Fewer hotspots overall, with SDG performance unevenly distributed, particularly for the elderly. New Territories: Exhibits the lowest concentration of high-SDG areas across all age groups, highlighting the need for focused interventions. Visualization and Interpretation: The heatmap gradient visually conveys SDG performance, with darker red areas indicating higher scores and lighter areas representing lower scores. By stratifying the data into age groups, the figure identifies demographic-specific inequalities and geographic disparities in sustainable development outcomes. This figure highlights the demographic nuances of sustainability in Hong Kong, emphasizing the need for targeted interventions to improve SDG outcomes for underserved age groups and regions. Policymakers can use these insights to prioritize resource allocation and enhance infrastructure and services for specific populations, particularly older residents in the New Territories and Hong Kong Island.

Figure 4 illustrates the spatial distribution of overall SDG scores across different age groups (<15, 15–39, 40–64, and >65) in three regions: Kowloon, Hong Kong Island, and New Territories. The maps use a heatmap-style gradient, where darker red areas indicate higher SDG scores, and lighter areas represent lower scores. This age group-based split aims to highlight how SDG outcomes vary for different age demographics within these regions. For instance, younger populations (<15 and 15–39) seem to experience relatively higher SDG scores in certain densely populated Kowloon and Hong Kong Island areas, as indicated by more concentrated red spots. In contrast, older populations (40–64 and >65) face more disparities, with fewer high-score areas visible across all three regions, particularly New Territories. By splitting the data into age groups, the figure visually communicates the uneven distribution of sustainable development outcomes, showing which age groups in each region may be underserved and where targeted interventions may be necessary to improve equity. The use of advanced geospatial models and analytical techniques enhances the precision of hotspot identification, providing a robust foundation for targeted interventions and policy formulations. The stratification by age divisions facilitates a more detailed exploration of the demographic dimensions influencing sustainability dynamics in different regions of Hong Kong.

Inequality metrics: Gini Coefficients of SDG progress across age groups and regions

Figure 5 illustrates a table of the Gini Coefficient corresponding to individual SDGs across various age groups and regions. The coefficients provide a quantitative measure of inequality in the distribution of progress toward achieving each SDG, offering insights into demographic and regional disparities within the framework of sustainable development.

Fig. 5: Gini coefficients of individual SDGs across different age groups and regions.

This figure presents a matrix of Gini coefficients for individual Sustainable Development Goals (SDGs) across various age groups and regions in Hong Kong. The Gini coefficient, ranging from 0 (perfect equality) to 1 (maximum inequality), is used here to quantify disparities in SDG progress, offering insights into the demographic and regional inequalities within the framework of sustainable development. Overall Inequality Observations: High Inequality: SDG 12 (Responsible Consumption and Production), SDG 14 (Life Below Water), and SDG 15 (Life on Land) exhibit Gini coefficients close to 0.99 across all age groups and regions, indicating significant disparities in achieving equitable progress for these SDGs. Moderate Inequality: Gini values for SDG 10 (Reduced Inequalities) and SDG 1 (No Poverty) show moderate inequality, particularly for older adults (65+), with notable regional disparities in Kowloon and Hong Kong Island. Age-Specific Inequality Patterns: Younger Populations (<15): Face substantial inequality in SDG 16 (Peace, Justice, and Strong Institutions), with Gini coefficients reaching up to 0.22 in New Territories and 0.21 in Kowloon, highlighting the challenge of ensuring inclusive societal structures for younger residents. Working-Age Groups (40–64): Exhibit relatively stable but moderately high inequality across SDG 8 (Decent Work) and SDG 3 (Good Health), suggesting gaps in economic and healthcare outcomes for this demographic. Elderly Populations (>65): Experience disparities in SDG 10 (Reduced Inequalities) and SDG 1 (No Poverty), with Gini coefficients exceeding 0.3 in some regions, reflecting vulnerabilities in income distribution and poverty reduction efforts. Regional Trends: Kowloon: Displays the most variability in inequality, with high Gini coefficients across multiple SDGs for all age groups. New Territories: Inequalities are pronounced in environmental SDGs, including SDG 13 (Climate Action), SDG 14, and SDG 15. Hong Kong Island: Shows moderate disparities overall but significant inequalities in SDG 12 and SDG 10 for older adults. Visualization and Interpretation: The heatmap uses a gradient from light pink (low inequality) to dark purple (high inequality) to visually represent the Gini coefficients across SDGs, age groups, and regions. Clusters of darker cells highlight areas where inequality is most pronounced, aiding in pinpointing critical gaps in progress. This figure underscores the uneven distribution of sustainable development progress across age groups and regions. It highlights the urgent need for age-specific and region-targeted policy interventions to address the disparities revealed by the Gini coefficients. Areas like SDG 12, SDG 14, and SDG 15 require significant attention, while younger and older populations in Kowloon and the New Territories emerge as key focus groups for achieving equitable SDG outcomes.

SDG 12 (Responsible Consumption and Production), SDG 14 (Life Below Water), and SDG 15 (Life on Land) show high Gini values (around 0.99) across all age groups and regions, indicating widespread inequality. This suggests that these areas face significant challenges in achieving equitable outcomes, regardless of region or age group. In contrast, younger age groups (under 15) tend to experience greater inequality in SDG 16 (Peace, Justice, and Strong Institutions), particularly in regions like Kowloon and New Territories, where Gini coefficients reach up to 0.21 and 0.22, respectively. Older adults (65+) also face disparities in areas like SDG 10 (Reduced Inequality) and SDG 1 (No Poverty), especially in Kowloon and Hong Kong Island. These insights suggest that targeted interventions are needed for regional and age-specific disparities to promote more equitable outcomes across the SDGs.

Validation and accuracy: comparing actual and calculated SDG scores in Hong Kong

Figure 6 presents a line chart showcasing the alignment between Actual Scores (AS) and Calculated Scores (CS), providing a visual representation of the validation and robustness of the methodology proposed in the research paper. The ‘Actual Scores’ refer to values calculated directly from data available on the United Nations’ official website, using a consistent methodology applied to the entirety of Hong Kong⁵².

Fig. 6: Line chart depicting the comparison between actual scores (AS) and calculated scores (CS), illustrating the validation and robustness of the proposed methodology.

Bar chart depicting the comparison between actual scores (AS) and calculated scores (CS), illustrating the validation and robustness of the proposed methodology.

The comparison between AS and CS based on SDG metrics offers insights into our methodology’s robustness. Most CS closely align with AS, indicating high accuracy and reliability. For instance, SDG2’s AS of 0.62 closely corresponds to the CS of 0.1 with a minor difference of 0.52. However, some deviations exist, like SDG17, where the AS is 2.21, and the CS is 0.3. Our focus on BLSS assessments ensures precision, tailored to individual contexts. Yet, some CS significantly differ from AS, such as SDG7, where the AS is 5.04, and the CS is 10. Despite discrepancies, our methodology offers nuanced assessments, enhancing understanding of sustainable development progress. In evaluating the accuracy of the aggregated building scores for Hong Kong, a t-test was conducted to compare the calculated scores to the actual ones. The null hypothesis (H₀) posited that there was minimal deviation between the aggregated and actual scores, implying that the aggregation method was reliable. The test resulted in a p-value of 0.146, higher than the typical significance threshold of 0.05, indicating insufficient evidence to reject the null hypothesis. Therefore, the results suggest that the calculated scores align reasonably well with the actual scores, supporting the accuracy of the aggregation process. Refer to Note 1 in the supplementary information for detailed calculations.

Scenario-based insights: optimizing facility configurations for enhanced community sustainability

The findings from this study facilitate the execution of scenario simulations, offering enhanced insights crucial for informed decision-making. Figure 7 delineates an original case study and the consequential scoring variations resulting from different decisions. In the baseline scenario, the residential score is lower for elderly care and childcare facilities, signaling a deficit in these specific amenities within the vicinity. In Scenario 1, prioritizing the establishment of an elderly care facility within a 5-km radius led to a score increase from 0 to 2.1, though the childcare score remained unaltered.

Fig. 7: Scenario analysis evaluating the impact of individual and integrated facility installations for informed decision-making.

This figure presents a scenario-based analysis of facility installations to optimize community sustainability scores. It evaluates the impacts of establishing individual versus integrated facilities for elderly care and childcare in a specific locality, highlighting changes in SDG-related service scores. Baseline Scenario: The original scenario represents the absence of nearby elderly care facilities (score = 0) and limited access to childcare facilities (score = 0.99). Other facilities, such as schools (score = 4.2), hospitals (score = 7.1), and police services (score = 5.1), are relatively well-established in the area. Scenario 1: Installation of an Individual Facility: In this scenario, a dedicated elderly care facility is introduced within a 5-km radius. Results: The elderly care score increases from 0 to 2.1, reflecting improved access to elderly care services. The childcare score remains unchanged at 0.99, as no interventions were made to improve childcare availability. Scenario 2: Installation of an Integrated Facility: This scenario introduces a combined elderly care and childcare facility, leveraging the benefits of shared services and intergenerational activities. Results: The elderly care score improves from 0 to 2.1. The childcare score significantly increases from 0.99 to 3, demonstrating the advantages of integrating services. Real-World Inspirations: Integrated facilities have been successfully implemented in global contexts, such as: Providence Mount St. Vincent (Seattle): Fostering intergenerational interactions between seniors and children, enhancing well-being. Guangzhou Yuexiu Social Welfare Center: Promoting daily interactions to improve social cohesion and care quality. Hong Kong Sheng Kung Hui Welfare Council: Combining elderly care, childcare, and family support for comprehensive community well-being. Visual Representation: Maps: The maps illustrate the spatial distribution of facilities and their corresponding scores under each scenario. Icons denote specific facility types, and score annotations provide a clear comparison across scenarios. Scenario Pathways: The flowchart-style layout visually distinguishes the baseline and modified scenarios, guiding the viewer through the decision-making process. Implications for Decision-Making: Scenario-based simulations highlight the importance of integrated facilities in addressing deficits in community services. This method provides a robust framework for policymakers to optimize facility configurations, balancing social compatibility, spatial feasibility, and service quality.

In Scenario 2, the framework advocates for an integrated facility, strategically combining elderly care and childcare services, considering their social and security compatibilities. Successful real-world implementations, such as the Providence Mount St. Vincent in Seattle and the Guangzhou Yuexiu Social Welfare Center, show the feasibility and benefits of integrated facilities. Providence Mount St. Vincent highlights intergenerational relationships through shared activities between seniors and children, enhancing their well-being. The Hong Kong Sheng Kung Hui Welfare Council provides integrated services, including elderly care, childcare, and family support, promoting intergenerational connections and improving the well-being of diverse community members through its comprehensive approach to care. Similarly, the Guangzhou Yuexiu Center promotes daily interactions that improve both groups’ social cohesion and care quality. The integrated facility implementation resulted in a substantial score improvement for both components. The elderly care facility score increased from 0 to 2.1, and the childcare facility score surged from 0.99 to 3. This nuanced scenario analysis provides a methodical basis for decision-makers, offering insights into optimal facility configurations and addressing specific deficiencies within the studied locality.

Discussion

In recent years, the global focus on sustainable development has intensified, guided by the United Nations’ SDGs^{53,54,55,56,57,58,59}. Research increasingly centers on assessing SDG implementation within the built environment, given buildings’ significant role in energy consumption, resource use, and emissions^{19,20,60,61,62,63,64,65,66}. As the 2030 deadline for the SDGs approaches, challenges arise despite their intentions^67,68. Urban development goals are complex and interconnected, varying between cities based on priorities like economic growth, social equity, and environmental sustainability. SDG 11: Sustainable Cities and Communities seeks to make cities inclusive, safe, resilient, and sustainable. However, rapid urbanization can lead to issues such as overcrowded housing, inadequate infrastructure, and poor access to essential services without proper urban planning and regulations. For instance, in Jakarta, Indonesia, rapid growth and poor infrastructure planning have led to overcrowded housing, poor sanitation, and severe traffic congestion. Low-income residents are pushed into flood-prone areas with limited services, while urban sprawl has increased deforestation and flood risks. Unplanned growth may exacerbate social inequalities by marginalizing low-income populations and reducing access to safe housing and public transportation. Additionally, along with deforestation, increased urban sprawl can contribute to environmental degradation through the loss of green spaces and higher pollution levels. Therefore, sustainable urban planning is essential to ensure cities develop in a way that balances social, economic, and environmental well-being for all residents. Traditional metrics often miss the nuances of urban dynamics and fail to track progress toward SDGs, such as neglecting environmental degradation or widening income gaps^69,70. This highlights the critical need for city-specific SDGs tailored to address the distinctive challenges of urban areas^71,72. This can be achieved by recognizing and addressing the specific challenges of each city and promoting equity and inclusivity⁷³.

The absence of validation in the existing SDG assessment system presents a significant challenge^74,75. This omission undermines the system’s credibility and effectiveness in measuring progress toward SDGs. Scenario simulations serve as invaluable tools in urban planning, offering the ability to forecast the impacts of constructing various facilities within a neighborhood⁷⁶. The ultimate goal is to create smart cities that adapt and thrive based on real-time data and projections. By integrating these simulations with digital twins—virtual replicas of the city itself—planners can assess the impact of new facilities in a safe, virtual environment^77,78,79,80. For instance, decision-makers can utilize simulations to model scenarios involving the introduction of amenities like parks, schools, or healthcare facilities, and assess their effects on community well-being, property values, and transportation dynamics⁸¹. Through the integration of demographic data, land-use patterns, and existing infrastructure information, simulations provide insights into how the addition of these facilities may enhance the livability and accessibility of the neighborhood⁸². Moreover, by considering different user personas, such as families, seniors, and commuters, urban planners can tailor recommendations to address specific needs and preferences⁸³. For instance, while young families may prioritize access to quality schools and recreational spaces, retirees may emphasize proximity to healthcare services and social amenities. Similarly, individuals with disabilities may require features like accessible housing and infrastructure to ensure mobility and independence. By incorporating the diverse requirements of these user personas into the planning process, urban planners can foster the development of inclusive and vibrant neighborhoods that cater to the needs of all residents⁸⁴. Similarly, SDG 5, which focuses on achieving gender equality and empowering all women and girls, underscores the importance of addressing the specific needs of different gender groups within urban development. In Hong Kong, the gender distribution is equitable, resulting in no significant differences observed based on gender (Refer to Fig. 1 in supplementary information). However, this may not be representative of global trends, making gender-based SDG scoring imperative to understand potential variations. Unfortunately, data in Hong Kong lacks gender specificity, hindering a comprehensive evaluation of SDG scores based on gender. Therefore, collecting gender-specific data is crucial for obtaining deeper insights into how SDG scores may vary among different demographic groups and ensuring more inclusive and targeted sustainability initiatives⁸⁵. However, Hong Kong stands out as an excellent example for this research due to its extensive data availability, surpassing that of many other cities at present.

This methodology applies to urban areas globally, although we have used Hong Kong as a specific example in our research. Hong Kong’s tax system heavily depends on revenue from land sales, which works hand in hand with its active real estate market to support urban sustainability and fairness. This research, using advanced geospatial models, uncovers key sustainability areas, guiding focused policymaking. By planning neighborhoods strategically to prioritize accessibility and essential amenities, governments attract real estate investment⁸⁶. At the same time, developers can demand that policymakers ensure buildings offer accessible facilities, creating attractive living spaces that meet diverse urban needs and maintain market appeal⁸⁷. However, the effectiveness of this approach relies on the availability of comprehensive and accurate datasets.

Urban planning and management can leverage the power of open data to promote transparency, collaboration, and data-driven decision-making⁸⁸. However, achieving this requires a balance between openness and privacy. Robust data governance frameworks are essential, employing anonymization techniques, secure encryption methods, and granular access controls⁸⁹. This ensures datasets empower innovation while upholding citizen trust. One crucial dataset, similar to the General Transit Feed Specification (GTFS) but tailored for cities, would provide standardized, real-time information on various urban elements⁹⁰. This could include transportation schedules, service disruptions, charging station availability and capacity, crowdsourced data, and even hospital wait times and bed vacancies. Standardized data formats (CSV, JSON, GeoJSON) and Application Programming Interfaces (APIs) would further streamline urban analysis by facilitating data integration and real-time updates across different systems and analytical tools⁹¹.

In the context of Hong Kong’s data-rich environment, insights from spatial analysis and machine learning techniques hold immense significance for urban planners, policymakers, and stakeholders. Urban planners stand to gain valuable insights into specific challenges and opportunities, enabling tailored strategies for both existing and new infrastructures and their impact on neighborhoods⁹². By understanding the unique characteristics of their urban environments, planners can tailor their strategies and interventions to address local needs more effectively⁹³. For example, if the research highlights a significant lack of green spaces in a particular neighborhood, planners can prioritize the development of parks and recreational areas to enhance residents’ quality of life and promote environmental sustainability. Policymakers can use the findings of this research to formulate evidence-based policies that promote economic prosperity, social equity, and environmental conservation⁹⁴. By leveraging insights into the interconnectedness of urban development goals, policymakers can develop comprehensive strategies that address multiple objectives simultaneously⁹⁵. For instance, if the research identifies a correlation between access to public transportation and economic opportunities, policymakers can invest in improving transit infrastructure to stimulate economic growth and reduce social disparities. Stakeholders, including community members, businesses, and advocacy groups, can benefit from this research by gaining a better understanding of the complex dynamics of urban development⁹⁶. By engaging with the research findings, stakeholders can contribute to more informed decision-making processes and advocate for policies and initiatives that align with their interests and values. For example, real estate developers can use the research to identify emerging trends and opportunities in urban markets, allowing them to make more strategic investment decisions and develop properties that meet the evolving needs of urban residents.

The strength of the methodology lies in its comprehensive approach, integrating geospatial models, machine learning algorithms, and extensive datasets. This framework bridges the gap between localized and citywide sustainability assessments, offering a nuanced understanding of sustainability dynamics across multiple levels, from individual buildings to regional landscapes. The engagement of diverse stakeholders becomes imperative in translating these insights into actionable policies that address the identified sustainability strengths and areas for enhancement. This research framework is suitable for urban areas worldwide, particularly in densely populated cities. This research methodology is adaptable for application in cities across various countries, including those in low- and middle-income regions. In our future work, we aim to extend this model to countries with limited data availability and lower granularity by using Open Street Maps data, which is open source and globally available. It effectively solves problems related to urban planning, infrastructure optimization, and inclusive growth. However, it may not be as applicable in rural areas and cannot predict all outcomes or address broader systemic issues. Future research could delve into the long-term impact of implemented interventions, providing insights into the sustainability trajectory over time.

Methods

This article introduces a novel methodology aimed at assessing SDGs at the building level, departing from conventional administrative unit evaluations. This approach, integrating advanced geospatial modeling and machine learning, enables granular analysis of individual buildings and projection of future scenarios, facilitating targeted interventions with maximum impact. Practical implementation demonstrates its potential for guiding urban planners, policymakers, and stakeholders in formulating strategic emissions reduction strategies, transcending one-size-fits-all approaches, and enriching understanding of emissions dynamics for more efficacious strategies in alignment with global sustainable development aspirations. Addressing this gap, the study develops an algorithm leveraging machine learning and geospatial models to map data points to the 17 SDGs and their 232 unique target indicators at the building level, yielding aggregated SDG scores for comparisons and actionable insights, applicable beyond Hong Kong to other cities and countries, thus offering a spatially informed approach for enhancing sustainability at various urban scales, fostering equitable and sustainable urban development through informed decision-making and policy formulation. This study utilizes a robust algorithm to collect live data efficiently and process the data in a desirable format, i.e., from spatial to tabular data. This data pipeline then uses algorithms to synthesize solutions to real-world scenarios. Figure 8 illustrates the methodology workflow.

Fig. 8: Comprehensive workflow illustration of the methodological process, integrating geospatial models and machine learning techniques for neighborhood sustainability assessment.

This figure illustrates the comprehensive workflow for assessing neighborhood sustainability using a combination of geospatial models and machine learning techniques. The methodological process integrates data collection, mathematical computation, and real-world applications to generate actionable insights for sustainable urban development. Data Collection and Processing: Step 1: Building Data (1.1) and Facilities Data (1.2) are collected as the foundational datasets. These datasets encompass information on building structures, proximity to services, and infrastructure. Step 2: Indicators relevant to each building are selected (2.1) and mapped to the Sustainable Development Goals (SDG) database (2.2) at the building level. This process ensures alignment with SDG indicators, facilitating granular and accurate analysis. Mathematical Calculation: Step 3: The shortest-path distance calculation determines the accessibility of facilities, contributing to SDG-related scores. Step 4: Individual SDG scores (SDG 1 to SDG 17) are computed, and a comprehensive Overall SDG Score is calculated as: ${\rm{Overall}}\; {\rm{SDG}}\; {\rm{Score}}=\mathop{\sum }\limits_{i=1}^{17}{X}_{i}$ Where ${X}_{i}$ represents the score for each SDG. Step 5: Scores are normalized and scaled for comparability, ensuring consistency across all indicators. Applications in Real-World Scenarios: Step 6: User preferences (e.g., priorities 1, 2, 3) are incorporated and weighted to generate optimal housing recommendations. Step 7: A machine learning-based housing recommendation system is developed, leveraging user inputs and SDG scores to suggest suitable housing options. Step 8: Gini coefficient calculations are applied to evaluate inequalities in SDG scores across hierarchical levels (building, district, region, and city), enabling targeted interventions. Visual Elements and Workflow Integration: Solid black arrows depict the integration of various stages, while dashed blue arrows represent the sequential steps in the process. Red arrows highlight applications and outputs relevant to real-world use cases, such as the recommendation system and inequality analysis. This workflow provides a robust framework for assessing and improving neighborhood sustainability. By integrating geospatial models and machine learning, it facilitates data-driven decision-making for urban planners and policymakers. The methodology offers scalable and actionable insights for enhancing sustainability outcomes at multiple hierarchical levels.

Collecting building and facilities data

The research methodology involves extracting spatial information from OpenStreetMaps and converting it into an organized Excel format. This foundational layer is enriched by integrating building-level Census Data, linking over 40,000 buildings to more than 5000 clusters. This process establishes a comprehensive spatial framework for urban analysis. Concurrently, diverse Facilities Data from 56 departments, comprising over 650 datasets, are harnessed. Extracted from the reputable Common Spatial Data Infrastructure (CSDI) portal⁹⁷, these data undergo precise geocoding before being integrated into layered maps. This dynamic approach provides valuable insights into urban infrastructure distribution and accessibility, enhancing the research’s analytical depth through authoritative open datasets.

Mapping building-level SDG indicators

The research methodology focuses on the process of mapping SDG indicators at the building level, ensuring alignment with urban sustainability objectives.

The process of indicator selection for this research aligns with the framework advanced by Butcher et al.⁹⁸. The framework delineates a structured approach for indicator identification, encompassing diverse granular levels, including the building level. A comprehensive data file provided by the aforementioned authors expounds upon the framework and delineates the meticulous process of indicator selection⁹⁹. It is worth noting that, in adherence to the framework, only datasets that were unavailable at the building level were omitted from consideration. Furthermore, this research strategy embraced an additional layer of inclusivity by incorporating datasets available at the building level, yet not explicitly mentioned within the framework. This judicious decision ensured the utmost inclusiveness of indicators considered when calculating holistic SDG scores. Figure 9 is a Sankey diagram illustrating the distribution of various indicators across the SDGs and their contribution to the final SDG score. Each left-side node represents a specific SDG indicator, while the middle nodes represent the SDGs to which these indicators contribute. Certain indicators (e.g., 8.5.2—Gender Pay Gap, or 3.b.3—Access to Affordable Medicines) contribute to multiple SDGs because they capture multidimensional aspects of sustainable urban development that transcend sectoral boundaries. Urban sustainability is inherently systemic, where improvements in one domain (such as employment equity or healthcare access) influence multiple development outcomes simultaneously.

Fig. 9: Illustrating the interplay of indicators and SDGs: highlighting interconnected measures and correlations.

This figure presents a Sankey diagram illustrating the distribution of indicators across various Sustainable Development Goals (SDGs) and their contribution to the final SDG score. The diagram visually represents the flow from specific indicators to SDG targets and, subsequently, to SDGs, offering insights into the structural composition of the SDG scoring process. Indicator-Target-SDG Flow: The diagram starts with specific indicators (e.g., 1.2.1, 8.5.2), which are grouped under corresponding SDG targets (e.g., 1.2, 8.5). Each target contributes to its respective SDG (e.g., SDG 1, SDG 8), culminating in the calculation of the final SDG score. Line Thickness and Indicator Distribution: The thickness of the lines connecting indicators, targets, and SDGs varies based on the number of indicators associated with each SDG. For instance: SDG 1 (No Poverty): Includes multiple indicators, resulting in thinner lines as the contributions are distributed across a broader range. SDG 2 (Zero Hunger): Has fewer indicators, leading to thicker lines, reflecting a more concentrated contribution. Structure of the SDGs: Each SDG is broken down into targets (e.g., 2.3, 3.b, 4.c), which are measurable objectives contributing to the SDG’s overall progress. Indicators (e.g., 2.3.2, 3.b.3) provide specific, quantifiable metrics used to assess progress toward each target. Purpose of the Diagram: The Sankey diagram visually emphasizes the distribution and density of indicators among SDGs without implying weight, significance, or priority. It highlights how some SDGs are supported by a broad set of indicators (e.g., SDG 1), while others are measured by a more focused set (e.g., SDG 9). Interpretation: The diagram demonstrates how the number of indicators affects the composition of SDG scores, showcasing the complexity and variability in measuring progress across different goals. It provides a transparent overview of the relationships between individual indicators, targets, and SDGs, facilitating better understanding and communication of the scoring process. This Sankey diagram offers a clear and structured visualization of how sustainability performance is assessed across SDGs, highlighting the intricate linkages between indicators, targets, and overall goals. It serves as a valuable tool for stakeholders to understand the multidimensional nature of SDG scoring and to identify areas where data collection and analysis may require further refinement.

For example, indicator 8.5.2, which measures gender pay disparities, not only pertains to SDG 8 (Decent Work and Economic Growth) by promoting equitable labor markets but also supports SDG 2 (Zero Hunger) by influencing income-driven food security outcomes, particularly among vulnerable populations. Similarly, indicator 3.b.3 on medicine accessibility directly advances SDG 3 (Good Health and Well-Being) but also indirectly bolsters SDG 10 (Reduced Inequalities) by reducing healthcare-driven socio-economic disparities.

Thematic overlaps arise because real-world urban challenges are interlinked, requiring cross-sectoral solutions rather than isolated interventions. Recognizing and modeling these interdependencies enhances the robustness of SDG assessments by capturing the cascading effects of individual indicators across multiple sustainability domains. This approach also mirrors the United Nations’ recognition that achieving the SDGs necessitates integrated action across economic, social, and environmental dimensions. A detailed breakdown of indicators and their parameter mappings is provided in Supplementary Table 1 for complete transparency.

This framework maps SDG indicators to census-derived statistics, focusing on occupation, income, age, and gender. Aligning SDG indicators with demographic data enables a detailed analysis of the connections between sustainable development goals and societal demographics and an understanding of complex relationships and trends. The methodology integrates SDG indicators with facilities data, including schools and healthcare establishments, enabling the analysis of accessibility and quantity at a granular, building-level perspective. Through the fusion of geospatial data and SDG metrics, comprehensive insights into the relationship between and the availability of essential services are obtained.

Shortest-path distance calculation from buildings to facilities

Utilizing Python libraries such as osmnx and networkx, the computation of street distances between buildings and nearby facilities was streamlined. This methodology integrated various modes of transportation, facilitating a comprehensive assessment of urban accessibility. Through this approach, the accessibility landscape was analyzed to identify areas with either strong or limited connectivity. The chosen centrality metric is based on enumerating the shortest pathways through a node, thus revealing its betweenness centrality. This pivotal measure is represented mathematically for a node denoted as i.

$${{CB}}_{i}=\sum _{a,b\epsilon i}\frac{\sigma (a,{b|i})}{\sigma (a,b)}$$

(1)

Where i is the set of nodes, σ(a,b) is the number of shortest(a,b)-paths, and σ(a,b|i) is the number of those paths passing through some node i other than a, b.

During the computation of the shortest distance, the pathway is segmented into discrete nodes at each juncture of directional alteration or transition between nodes. The cumulative distance is derived by aggregating these inter-node distances.

$$\begin{array}{ll}d=\sqrt{{({x}_{12}-{x}_{11})}^{2}+{({y}_{12}-{y}_{11})}^{2}}+\sqrt{{({x}_{22}-{x}_{21})}^{2}+{({y}_{22}-{y}_{21})}^{2}}+\ldots \ldots .\\ \,+\sqrt{{({x}_{n2}-{x}_{n1})}^{2}+{({y}_{n2}-{y}_{n1})}^{2}}\end{array}$$

(2)

Where d is the total shortest distance.

${x}_{n1},\,{y}_{n1}$ are the coordinates of ‘${n}^{{th}}$’ node.

Computation of individual and overall SDG scores

Building-level SDG scores were computed through a multi-step process designed to ensure transparency, comparability, and replicability. First, each building was associated with multiple SDG indicators based on its spatial, demographic, and facility attributes. For each relevant indicator, a raw value was extracted or computed, such as shortest-path distances to essential facilities, building population characteristics, and accessibility measures.

Each individual indicator score was first normalized using min-max scaling to a 0–10 scale, ensuring that all indicators, regardless of their original units, could be meaningfully compared across buildings:

$$Scaled\,Score\,\left(S\right)=\frac{S-Min(S)}{Max\left(S\right)-Min(S)}\times 10$$

(3)

Where S is the raw indicator value.

Next, for each building, normalized scores of all indicators corresponding to a particular SDG were summed to derive the building’s score for that SDG:

$${SDG}_{i}{Score}_{b}=\displaystyle \mathop{\sum }\limits_{k=1}^{n}Scaled\,Indicator\,{Score}_{k,b}$$

(4)

Where n is the number of indicators mapped to SDG i for building b.

Finally, a cumulative overall SDG score was calculated by summing all individual SDG scores for that building:

$${Overall\,SDG\,Score}_{b}=\displaystyle \mathop{\sum }\limits_{i=1}^{17}{SDG}_{i}\,{Score}_{b}$$

(5)

Figure 10 shows the table listing how each indicator is calculated based on the corresponding indicator.

Fig. 10: Mapping of Sustainable Development Goals (SDGs) to specific indicators and the corresponding parameters used for building-level analysis.

Each SDG score is computed by aggregating its constituent indicators, which are derived from census variables, facility accessibility metrics, and urban infrastructure data. The parameters listed represent the operationalized variables used to quantify each indicator within the geospatial modeling framework. This mapping ensures a transparent and standardized methodology for evaluating multi-dimensional urban sustainability at the building scale. Mapping of SDG indicators and their associated parameters used for building-level sustainability assessment. Each Sustainable Development Goal (SDG) is broken down into specific target indicators (e.g., 1.1.1, 8.5.2), with corresponding parameters derived from census data and facility accessibility metrics. Parameters include socio-economic measures (e.g., household income, education levels), infrastructure accessibility (e.g., access to public transport, healthcare facilities), and environmental metrics (e.g., recycling rates, resilience to climate hazards). Some indicators inform multiple SDGs due to thematic overlaps, reflecting the interconnected nature of sustainable urban development. This mapping underpins the computation of individual building SDG scores and provides the foundation for the holistic assessment framework adopted in this study. A detailed description of parameter derivation and indicator calculation is available in the Supplementary Information.

Equal weighting was applied to all indicators within an SDG and across SDGs, ensuring that no single goal or indicator was prioritized unless specified by data availability. This choice was made to maintain neutrality and comparability across buildings and locations. A full breakdown of indicator-SDG mappings and corresponding weights is provided in Table 1 of the Supplementary Materials. A computational pipeline and detailed Python code for SDG 1 and its indicators (1.1.1, 1.2.1, 1.4.1, 1.a.2) are also provided in the Supplementary File, refer to Note 2, demonstrating how these scores were calculated.

Incorporating and weighting user inputs to generate optimal housing recommendations

The methodology encompasses the collection of user inputs, encompassing personal identifiers like name and age, financial aspects such as income and housing budget, as well as geospatial preferences including desired location and prioritized proximate facilities. These inputs serve as pivotal criteria in generating housing recommendations that align with individual preferences. Through a systematic process, the collected inputs are weighted according to their respective priority levels. Employing a comprehensive analytical framework, the system generates and presents the most suitable housing suggestions, integrating user preferences with quantifiable criteria for informed decision-making.

The methodology involves the solicitation of user-defined facility preferences, wherein the significance of each preference is determined by its assigned priority level. This hierarchical structure ensures that the most critical facility preference receives the highest weightage in subsequent analyses.

Mathematically, the process can be represented as follows:

Let P be the set of user-defined facility preferences, and ${W}_{p}$ denote the weight assigned to each preference p based on its priority level. The weightage normalization can be expressed as:

$$\sum _{p\in P}{W}_{p}=1$$

(6)

Furthermore, ${W}_{p}$ influences the evaluation of facility suitability scores, such that higher weights are accorded to preferences with elevated priority levels. This adjustment is formulated as:

$${{Adjusted\; Facility\; Score}}_{p}={{Original\; Facility\; Score}}_{p}\times {W}_{p}$$

(7)

This systematic approach guarantees that the user’s most significant facility preference exerts a substantial influence on the final housing recommendation, thereby enhancing the personalized and relevant nature of the outcome.

Developing a machine learning-based housing recommendation system

Upon the acquisition of user information, a comprehensive user database is established, wherein individual attributes such as age, income, and occupation are stored. This repository serves as the foundation for identifying users with analogous interests and characteristics. Additionally, the database logs user interactions, capturing their interactions with suggested housing options and gauging their engagement levels. This integration of attributes is utilized to develop a recommendation system utilizing cosine similarity. This mechanism orchestrates the presentation of housing recommendations to users who share resemblances in their preferences and characteristics.

Mathematically, the process can be articulated as follows:

Let U represent the set of users, ${A}_{u}$ signify the attribute vector of user u, and ${C}_{u}$ denote the vector of user clicks on housing suggestions. The user database then encompasses (U,A,C) tuples. The cosine similarity calculation for user u with v is expressed as:

$${{Cosine\; Similarity}}_{u,v}=\frac{{A}_{u}\times {A}_{v}}{\left|\left|{A}_{u}\right|\right|\times \left|\left|{A}_{v}\right|\right|}$$

(8)

Based on the derived cosine similarity scores, users with comparable attributes and interaction patterns are identified. Consequently, housing recommendations aligned with these commonalities are presented, enhancing the system’s ability to offer relevant and tailored suggestions to users.

Gini Coefficient calculation of SDG scores across multiple hierarchical levels: building, district, region, and city

After the computation of SDG scores across multiple levels, such as building, district, regional, and overall city. The Gini coefficients are determined individually for each SDG, in addition to an aggregate Gini coefficient for the overall SDG score. These coefficients serve as quantitative metrics to assess disparities and inequalities in the granularity of individual buildings.

Mathematically, for a given set of SDG scores $S=\{{s}_{1},\,{s}_{2},\,\ldots ,\,{s}_{n}\}c$ where $n$ represents the number of buildings, the Gini coefficient $G$ is calculated as:

$$G=\frac{2{\sum }_{j=1}^{n}{\sum }_{j=1}^{n}|{s}_{i}-{s}_{j}|}{n{\sum }_{i=1}^{n}{s}_{i}}$$

(9)

Where ${s}_{i}$ and ${s}_{j}$ denote the SDG scores of buildings i and j, respectively.

The derived Gini coefficients offer insights into the distributional patterns of SDG scores across buildings, unveiling potential imbalances in the attainment of. The individual Gini coefficients per SDG and the overall Gini coefficient provide a comprehensive picture of inequities, enriching the understanding of localized disparities and contributing to informed policy interventions.

Statistical analysis of actual and calculated score: paired t-test

A step-by-step explanation and results for a paired t-test is performed to determine whether there is a significant difference between the Actual and Calculated SDG scores. Null Hypothesis (H₀): There is no significant difference between the Actual Scores and Calculated Scores. The mean difference is zero.

$${H}_{0}:{\mu }_{{diff}}=0$$

(10)

Alternative Hypothesis (H₁): There is a significant difference between the Actual Scores and Calculated Scores. The mean difference is not zero.

$${H}_{1}:{\mu }_{{diff}}\,\ne \,0$$

(11)

Next, the standard deviation is calculated using the following formula.

$$s=\sqrt{\frac{1}{n-1}\sum {\left({x}_{i}-\bar{x}\right)}^{2}}$$

(12)

The paired t-test is performed and further compared to the critical value for a given significance level (α = 0.05).

$$t=\frac{\bar{d}}{{s}_{d}/\sqrt{n}}$$

(13)