MACHINE LEARNING IS APPLIED IN PREDICTING THE LIKELIHOOD OF A TRAFFIC COLLISION

Road accidents remain a leading cause of death, disability, and hospitalization nationwide, underscoring the critical need for effective traffic accident risk prediction to save lives. Various models, from traditional statistical methods to modern machine learning approaches, have been proposed for this purpose. This study compares these models to identify effective strategies for predicting traffic accident risks. Given that drivers bear responsibility on the road, the research aims to provide predictive insights based on factors drivers can anticipate, such as vehicle type, age, gender, time of day, weather conditions, and more. Models like Random Forest, Logistic Regression, and Optimal Classification Trees are explored, aiming for intuitive outcomes beneficial to drivers. Additionally, geo-location data analysis using K-means clustering offers insights into high-accident areas. As the number of vehicles increases, managing road design and traffic becomes more challenging. Globally, road accidents pose significant concerns due to their profound impact on safety, health, and well-being. The World Health Organization (WHO) estimates that 1.35 million people die annually in road accidents, driving extensive research into advanced methodologies and algorithms for prediction and analysis. While external factors contribute to many accidents, driver-related factors also play a crucial role. Adverse weather conditions, such as rain, clouds, and fog, often impair visibility and increase accident risks. The current predictive model evaluates several potential causal factors in this context.