Crash comparison of vehicles with modern safety features versus conventional vehicles using Artificial Intelligence: a deep learning-based analysis

As automotive technology advances, modern vehicles are becoming increasingly integrated with advanced safety systems designed to prevent collisions or minimize injuries. Still, their effectiveness compared to older cars is not well documented, particularly in developing regions. This research highlights the gap by examining the differences in crash outcomes between vehicles equipped with sophisticated safety features, such as Adaptive Cruise Control (ACC), Automatic Emergency Braking (AEB), and Lane Departure Warning (LDW), and those without these features. We obtained traffic and insurance crash data from Jordan, which were supplemented with demographic, contextual, and vehicle-related information, then predicted, balanced, and enriched. A hybrid deep learning model, based on Convolutional Neural Networks (CNN) and Long Short-Term Memory (LSTM) networks, is used for the analysis. Four deep learning models, 1D CNN, RNN, LSTM, and CNN-LSTM, were evaluated for their ability to classify injury severity levels. The CNN-LSTM model consistently outperformed others, achieving an accuracy of 97% and an AUC of 99%. These results highlight the effectiveness of hybrid models in capturing both spatial and temporal patterns in crash data. Furthermore, our findings demonstrate that modern safety-equipped vehicles are associated with significantly lower injury severity, supporting the case for broader adoption of these technologies. The study not only contributes to the growing body of research on AI in traffic safety but also offers valuable insights for policymakers, car manufacturers, and urban planners aiming to enhance road safety through data-driven strategies.