Development of Naive ML-based Crash Prediction for Transport Safety

The rapid escalation of global vehicular traffic has necessitated a shift from reactive to proactive road safety management. Traditional statistical methods often struggle to capture the non-linear and complex relationships inherent in traffic accident data, which is frequently characterized by high dimensionality and class imbalance. This research focuses on the development of a machine learning-based framework designed for crash prediction and impact analysis to enhance transport safety. By utilizing historical accident records, real-time traffic flow data, and meteorological conditions, the study implements a suite of algorithms including Naive Bayes, Random Forest, and Extreme Gradient Boosting (XGBoost) to predict the likelihood and severity of collisions. A significant portion of the study is dedicated to impact analysis, identifying the critical factors that contribute most significantly to accident outcomes. The proposed methodology emphasizes data preprocessing techniques, including the Synthetic Minority Over-sampling Technique (SMOTE), to address the rarity of fatal crashes in datasets. The expected outcomes suggest that the integrated machine learning approach can significantly outperform traditional logistic regression models in both sensitivity and precision. By providing high-accuracy predictive insights, this research offers a robust tool for traffic authorities to implement real-time interventions and reducing the socio-economic burden of road traffic injuries.