Profiling Driver Behaviors Using AI-Based Methods and Deep Learning Techniques for Improving Road Safety: A Comparative Study of Algorithms

Traffic accidents remain a significant global issue, resulting in substantial human and economic losses annually. This study focuses on analyzing and profiling driver behaviors to enhance traffic safety and understand the causes of accidents. Utilizing a comprehensive dataset that includes a diverse set of real-world driving metrics, including velocity, acceleration, and braking behavior, along with simulated driving scenarios, we leverage both machine learning and deep neural network approaches to classify driver behaviors. Our methodology incorporates preprocessing, feature selection and model training using algorithms like XGBoost, CNN+BiLSTM, Decision Forest, LSTM, RNN and LSTM+XGBoost. The study demonstrates the superior performance of XGBoost in profiling driver behaviors, achieving the highest accuracy among the models tested. The findings offer valuable insights for developing strategies to improve traffic safety, inform driver training programs and shape policies aimed at reducing traffic accidents. Future work will focus on real-time continuous monitoring of driver actions in real time and the incorporation of cutting-edge driver support technologies to promote safer driving practices.