Analysis of Road Traffic Crash Severity Through Machine Learning on Indian Highways

• Published in: Journal of the Eastern Asia Society for Transportation Studies Vol. 15; pp. 3228 - 3244
• Main Authors: GARG, Tanishq, CHATTERJEE, Sudipa, MAITRA, Bhargab
• Format: Journal Article
• Language: English
• Published: Eastern Asia Society for Transportation Studies 2024
• Subjects: artificial neural network; crash severity; feature importance; hierarchical clustering; low- and middle-income countries
• ISSN: 1881-1124
• DOI: 10.11175/easts.15.3228

Fatal crash comes with a huge cost of person loss besides vehicle, goods, and property damage which impacts the overall economy of the country. Therefore, there is a need to identify factors responsible for the high severity of crashes, especially for low and middle countries. Numerous studies have already proved that Machine Learning techniques are more effective than traditional statistical modelling and have recently been extensively used in road safety research. The present study is motivated to study the effect of various parameters on road crash severity and derive measures to enhance safety using exploratory analysis, neural network, and unsupervised agglomerative clustering. The insights obtained from the study indicated that the time of the crash, speed limit signs, and road type as the top three features. The insights obtained from the study were utilized to propose recommendations for safety improvement on the Indian road network.