The Sustainable Development Goals (SDGs), adopted by the United Nations, aim to achieve global social, economic, and environmental well-being. Cities play a crucial role in implementing the SDGs, as approximately 65% of the 169 targets require city engagement. While national governments demonstrate commitment, rapid urban development presents challenges like inequality, pollution, and infrastructure deficits. Municipal governments are integrating SDGs into their development plans, but quantitative assessments of SDG progress at the city level, especially for small and medium-sized cities in developing countries, are limited due to data scarcity. Existing quantitative assessments of SDG progress have been conducted at global, regional, national, and subnational levels. The SDG Index score (arithmetic mean of 17 individual SDG scores) is useful for comparing overall SDG performance. However, large-scale sustainability assessments for all cities of varying sizes remain limited, particularly in developing countries where relevant statistical data are scarce. This data shortage significantly hinders the development of holistic strategies to promote city sustainability. The abundance of big data offers an opportunity to address this data gap. Big data's characteristics—large volume, high velocity, variety, veracity, and value—enable timely and efficient monitoring of SDG progress. Several studies have utilized various types of big data (nighttime light imagery, point-of-interest data, OpenStreetMap data) to monitor specific SDG indicators. However, these studies often focus on individual or a few indicators within a specific SDG, lacking a holistic assessment of multiple SDGs. This study addresses this gap by developing a systematic method to assess overall SDG progress for cities of varying sizes. The researchers constructed a generic indicator system using open-source big data and an artificial neural network (ANN) model to evaluate the SDG Index for 254 Chinese cities with sufficient statistical data and then applied the model to assess 83 cities with limited data. This comprehensive evaluation identifies priorities for city-level SDG implementation.

The literature review highlights the importance of measuring and assessing progress towards the SDGs in various contexts. Existing studies on SDG assessments have been conducted at different scales (global, regional, national, and subnational), utilizing various methodologies and indicator frameworks. The SDG Index score, calculated as the arithmetic mean of 17 individual SDG scores, has been widely used as a useful tool to compare the overall SDG performance of different entities. However, existing studies focusing on city-level SDG assessments are limited, especially for smaller cities in developing countries. The lack of comprehensive and readily available statistical data at the city level poses a significant challenge. The literature also points to the emerging use of big data as a promising alternative for monitoring SDG progress. While studies have explored the use of big data for tracking specific indicators of particular SDGs, a gap exists in comprehensively integrating multi-source big data for a holistic SDG assessment across various city sizes and contexts.

The study employed a two-pronged approach: first, assessing SDG performance using statistical data for 254 Chinese cities and then utilizing big data and an ANN model for a more comprehensive assessment, including cities lacking statistical data. **Calculation of Individual SDG and SDG Index Scores Using Statistical Data:** The researchers developed a localized indicator framework, selecting 54 indicators across 17 SDGs based on data availability and consistency in statistical caliber at the city level. These indicators were normalized to a 0-100 scale, accounting for upper and lower bounds based on best and worst performance, respectively. Equal weights were assigned to each SDG and its indicators. Individual SDG scores and overall SDG Index scores were calculated using weighted averages. **Calculation of SDG Index Scores Using Multisource Big Data:** The study leveraged open-source big data, including remote sensing data (nighttime light imagery, land use data), and geospatial big data (point-of-interest data, company information, gridded population data, road networks). Correlation analysis was used to select relevant big data indicators for each SDG, ensuring high correlation with the corresponding statistical data indicators. An ANN model, specifically a backpropagation (BP) neural network, was trained using the 18 selected big data indicators as input and the SDG Index scores from the statistical data as output. The model was trained using 70% of the data, validated with 15%, and tested independently with the remaining 15%. The Levenberg-Marquardt algorithm was used for network training. Finally, the trained ANN model was used to predict SDG Index scores for the 83 cities lacking sufficient statistical data. The data used in the study includes statistical data from national population surveys, city-level statistical communiques, city statistical yearbooks, carbon emission data, government performance data, marine data, company information, point-of-interest (POI) data, road network data, population data from WorldPop, land use data, and nighttime light data.

The study revealed several key findings: * **City-level Individual SDG Performance:** Analysis of 254 cities using statistical data indicated variations in SDG performance across city size and location. While large cities generally performed better across most SDGs, they faced challenges in reducing income inequality (SDG 10) and increasing global partnerships (SDG 17). Small and medium-sized cities performed relatively better on SDG 15 (life on land), but still lagged significantly. * **Regional Differences:** Eastern cities performed better across most SDGs, due partly to economic development and foreign investment. Central cities showed better energy consumption performance, while northeastern cities performed highly on SDG 2 (zero hunger) because of fertile agricultural land. Western cities lagged due to limited ecological assets, inadequate educational facilities, and lower levels of external communication. * **Harnessing Big Data:** The correlation analysis showed strong relationships between the 18 selected big data indicators and their corresponding SDG scores. The ANN model demonstrated good predictive performance, with a low root-mean-square error (RMSE = 3.13) and a high coefficient of determination (R² = 0.7625) for the test set. This indicates the model's ability to accurately estimate SDG Index scores using only big data indicators. * **City-Level SDG Index Performance:** The spatial distribution analysis using the ANN-predicted SDG Index scores for all 337 cities (including the 83 with insufficient statistical data) showed higher scores in Eastern China and lower scores in western cities. Provincial capital cities generally exhibited higher scores than other cities within the same province. The analysis also showed that the SDG Index scores decreased with decreasing city size and income level. * **Comparison of Statistical Data and Big Data Results:** Although there was a high correlation between the SDG Index scores calculated from statistical data and those predicted using the ANN model, the study highlighted the efficiency and cost-effectiveness of using big data. The big data approach offers a more timely and readily available assessment than the lag associated with official statistics.

The findings address the research question by demonstrating the feasibility and value of using big data for assessing SDG progress in cities of varying sizes. The study successfully developed and validated an ANN model that accurately predicts SDG Index scores using readily available, low-cost big data. This overcomes the limitations of relying solely on traditional statistical data, which are often unavailable or incomplete for smaller cities in developing regions. The consistent trend of decreasing SDG Index scores with decreasing city size highlights the need for targeted interventions to improve sustainability in smaller cities. The regional disparities observed emphasize the importance of context-specific policies and resource allocation. The results contribute to the field by providing a robust and replicable methodology for assessing city-level SDG progress. The use of open-source big data promotes accessibility and expands the scope of SDG monitoring beyond data-rich areas. The findings underscore the need for a more balanced approach to sustainable development, acknowledging the unique challenges and opportunities faced by cities of different sizes and locations.

This study successfully developed and validated an ANN model using multi-source, low-cost big data to efficiently assess the overall SDG performance of cities of varying sizes. The results highlight the importance of big data for SDG monitoring, particularly in data-scarce settings. The findings show significant variations in SDG performance across city sizes, locations, and income levels, with smaller cities generally lagging behind. The study advocates for creating integrated socioeconomic communities to foster the development of smaller cities and achieve balanced national sustainability. Future research should focus on increasing investments in SDG data and evaluation systems and exploring the synergies and trade-offs between different SDGs at the city level.

The study's limitations include the reliance on a specific set of big data indicators and their correlation with existing statistical data. While the ANN model demonstrated strong performance, the accuracy could be improved by incorporating more diverse data sources and refining indicator selection. The study focused on Chinese cities; therefore, the generalizability of the findings to other contexts might be limited. Additionally, the study focused primarily on the overall SDG Index, potentially overlooking nuanced insights from individual SDG targets.