Predicting and explaining severity of road accident using artificial intelligence techniques, SHAP and feature analysis

• Published in: International journal of crashworthiness Vol. 28; no. 2; pp. 186 - 201
• Main Authors: Panda, Chakradhara, Mishra, Alok Kumar, Dash, Aruna Kumar, Nawab, Hedaytullah
• Format: Journal Article
• Language: English
• Published: Cambridge Taylor & Francis 04.03.2023; Taylor & Francis Ltd
• Subjects: Accident data; Accident prediction; Accident severity prediction; AI-enabled accident management system; Artificial intelligence; Commercial vehicles; feature analysis; Literature reviews; machine learning; Pedestrians; road safety; Roads & highways; SHAP; Traffic accidents
• ISSN: 1358-8265; 1754-2111
• DOI: 10.1080/13588265.2022.2074643

The accurate prediction of accident severity has become an active area of research in recent years, although studies in certain regions such as South Asia and Sub-Saharan Africa are comparatively less. In this study, we aim to contribute in many ways: (i) we conduct an analytical review of the literature to gauge the interest and scope of existing studies and identify the direction for further research, and (ii) a mixture of old and relatively new artificial intelligence (AI) techniques is applied to road accident data of India (iii) we employ shapley additive explanations (SHAP) for interpretation of AI model predictions, and (iv) an AI-enabled accident management system is proposed. The findings suggest that AI models are capable of predicting the accident severity. Precisely, the gradient boosting machine attains the best test accuracy. Among features, commercial vehicles, excess speed, national highways, and pedestrians' fault are responsible for accidental road killings.