The automotive industry has achieved remarkable advances in passenger car safety systems to mitigate the risk of injuries and fatalities caused by road accidents. However, to further improve vehicle safety, it is essential to have a deeper understanding of real-world accidents and the true safety benefits of various safety systems in the field. This requires a framework to evaluate the effectiveness of safety systems in reducing occupant injury and fatalities.
This study aims to use machine-learning techniques to predict occupant injury severity by considering accident parameters and safety systems, using the Road Accident Sampling System - India (RASSI) real-world accident data. The RASSI database contains comprehensive accident data, including various factors that contribute to occupant injury. The study focused on fifteen accident parameters that represent key aspects of crash scenarios such as vehicle type, accident type, vehicle speed, and occupant details. Multiple machine learning algorithms such as decision tree, random forest, and neural network are applied to build robust models for predicting injury severity. The performance of each machine-learning model is assessed using appropriate metrics such as accuracy, precision, recall, and F1 score. Furthermore, a feature importance analysis is performed to identify the critical factors that influence the injury severity.
The results show the effectiveness of the proposed approach in accurately predicting occupant injury severity across different crash scenarios. Moreover, the study provides an opportunity to gain valuable insights into the underlying factors that affect occupant injury severity. This will help safety engineers to conduct studies to understand the effectiveness of safety systems independently. This will assist selection and prioritization of various safety systems towards enhancing occupant safety considering real-world accident scenarios.