A Data-Driven Assessment and Prioritization of Urban Road Accident Burden in Indian Cities for Transportation Safety

Abstract

Urban road-safety planning requires a transparent basis for identifying where engineering, enforcement and public-awareness resources may be prioritized. This study develops a reproducible framework to assess reported road accident burden across 50 Indian million-plus cities using official statistics. The 2023 city-level dataset contains 81,144 reported accidents, 17,255 fatalities and 73,476 injuries. The framework combines aggregate trend analysis, entropy-weighted Technique for Order Preference by Similarity to Ideal Solution (TOPSIS), K-means clustering, ranking sensitivity analysis and a two-axis burden–severity matrix. Entropy weighting assigns weights of 0.3412, 0.2695 and 0.3894 to accidents, fatalities and injuries, respectively. Delhi, Bengaluru and Jabalpur are the three highest-ranked cities under the composite burden score. A five-cluster solution yields a silhouette coefficient of 0.4387 and mean pairwise adjusted Rand index of 0.9929 across 20 initializations. Rank agreement is high under equal, accident-focused and injury-focused weights, but lower when fatalities receive dominant emphasis. The results show that city prioritization depends on the selected outcome dimension. The framework is a screening-level decision-support approach: because consistent city-specific exposure denominators are unavailable, the results represent reported burden and must not be interpreted as exposure-adjusted crash risk or causal evidence.

Introduction

This study analyses road traffic accidents in 50 major Indian cities using data from the Ministry of Road Transport and Highways (MoRTH) for 2019–2023. It aims to identify cities with higher reported accident burdens, compare fatalities and injuries, classify cities into different burden profiles, and examine how rankings change under different evaluation criteria.

The study uses three main indicators: total accidents, fatalities, and injuries. An entropy-weighted TOPSIS method is applied to rank cities, while K-means clustering groups them into five descriptive profiles. Sensitivity analysis tests the effect of different indicator weights, and a separate burden–severity matrix distinguishes accident volume from fatality severity.

The results show that accidents and fatalities decreased significantly in 2020, likely due to COVID-19 mobility restrictions, before rising again towards 2019 levels by 2023. Under entropy-weighted TOPSIS, Delhi ranked first, followed by Bengaluru, Jabalpur, Chennai, and Malappuram. The five-cluster analysis achieved a silhouette score of 0.4387 and high clustering stability, with an average adjusted Rand index of 0.9929.

The study highlights that cities with more accidents do not necessarily have higher fatality severity. It also emphasizes that accident counts alone cannot measure actual road safety risk because factors such as population, traffic volume, and vehicle-kilometres travelled are not included.

Conclusion

A data-driven framework was developed to assess reported road accident burden across 50 Indian million-plus cities using official statistics. Entropy-weighted TOPSIS, K-means clustering, sensitivity analysis and a burden–severity matrix provide complementary summaries of accident, fatality and injury counts. Delhi, Bengaluru and Jabalpur were the highest-ranked cities under the composite burden score. The five-cluster solution was stable across the tested initializations, while sensitivity analysis showed that the ranking changed more substantially when fatalities were given a dominant weight. The principal value of the framework is a transparent screening procedure that combines multiple indicators without conflating different geographic aggregations or overstating what the data can establish. Its outputs can help direct subsequent data collection and local safety assessment, but they should not replace exposure-based risk modelling, road-safety audits or site-level engineering investigation.

References

[1] World Health Organization. (2023). Global status report on road safety 2023. Geneva: WHO. https://www.who.int/publications/i/item/9789240086517 [2] World Health Organization. (2021). Global plan for the decade of action for road safety 2021–2030. Geneva: WHO. [3] Ministry of Road Transport and Highways. (2025). Road accidents in India 2023. Transport Research Wing, Government of India. https://morth.nic.in/sites/default/files/Road-Accident-in-India-2023-Publications.pdf [4] World Health Organization. (2010). Data systems: A road safety manual for decision-makers and practitioners. Geneva: WHO. [5] Barffour, M., Gupta, S., Gururaj, G., & Hyder, A. A. (2012). Evidence-based road safety practice in India: Assessment of the adequacy of publicly available data in meeting requirements for comprehensive road safety data systems. Traffic Injury Prevention, 13(Suppl. 1), 17–23. https://doi.org/10.1080/15389588.2011.636780 [6] Mohan, D., Tiwari, G., & Mukherjee, S. (2016). Urban traffic safety assessment: A case study of six Indian cities. IATSS Research, 39(2), 95–101. https://doi.org/10.1016/j.iatssr.2016.02.001 [7] Paul, A. B. (2024). Investigating and modeling motorcycle crash rates at urban unsignalized intersections in India. Traffic Injury Prevention, 25(6), 852–859. https://doi.org/10.1080/15389588.2024.2348729 [8] Anderson, T. K. (2009). Kernel density estimation and K-means clustering to profile road accident hotspots. Accident Analysis & Prevention, 41(3), 359–364. https://doi.org/10.1016/j.aap.2008.12.014 [9] Maji, A., Velaga, N. R., & Urie, Y. (2018). Hierarchical clustering analysis framework of mutually exclusive crash causation parameters for regional road safety strategies. International Journal of Injury Control and Safety Promotion, 25(3), 257–271. https://doi.org/10.1080/17457300.2017.1416485 [10] Hwang, C. L., & Yoon, K. (1981). Multiple attribute decision making: Methods and applications. Springer. [11] Shannon, C. E. (1948). A mathematical theory of communication. Bell System Technical Journal, 27(3), 379–423; 27(4), 623–656. [12] MacQueen, J. (1967). Some methods for classification and analysis of multivariate observations. In Proceedings of the Fifth Berkeley Symposium on Mathematical Statistics and Probability (Vol. 1, pp. 281–297). University of California Press. [13] Rousseeuw, P. J. (1987). Silhouettes: A graphical aid to the interpretation and validation of cluster analysis. Journal of Computational and Applied Mathematics, 20, 53–65. https://doi.org/10.1016/0377-0427(87)90125-7 [14] Hubert, L., & Arabie, P. (1985). Comparing partitions. Journal of Classification, 2, 193–218. https://doi.org/10.1007/BF01908075 [15] Papadimitriou, E., Yannis, G., Bijleveld, F., & Cardoso, J. L. (2013). Exposure data and risk indicators for safety performance assessment in Europe. Accident Analysis & Prevention, 60, 371–383. https://doi.org/10.1016/j.aap.2013.04.040

Copyright

Copyright © 2026 Macha Anantha Lakshmi, B. Vinod Kumar, T. Kiran Kumar. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.