A Detailed Review on Machine Learning-Based Crash Prediction and Impact Analysis for Transport Safety

Road traffic accidents remain a significant global concern, leading to substantial loss of life, injuries, and economic damage. With the increasing complexity of transportation systems, traditional statistical methods have proven inadequate in capturing the dynamic and nonlinear interactions among factors influencing crash occurrences. In recent years, machine learning techniques have emerged as powerful tools for crash prediction and impact analysis due to their ability to process large-scale, heterogeneous datasets and identify hidden patterns. The study examines various machine learning models, including traditional algorithms such as Logistic Regression, Naïve Bayes, Decision Trees, and advanced methods such as Random Forest, Gradient Boosting, and deep learning techniques. It highlights the role of multi-source data integration, including traffic flow, weather conditions, road infrastructure, and driver behavior, in improving prediction accuracy. The review also explores the challenges associated with data imbalance, model interpretability, and real-time implementation. The expected outcome of this review is to provide a comprehensive understanding of current methodologies and guide future research toward developing efficient, interpretable, and scalable solutions for transport safety.