MACHINE LEARNING IN TRAFFIC SAFETY: A REVIEW OF TECHNIQUES FOR INJURY SEVERITY PREDICTION

Abstract

Vehicle crashes are a significant cause of fatalities and serious injuries annually, presenting an ongoing public health problem on a global scale. In order to effectively address specific solutions, improve emergency response, and elevate road safety regulations, researchers must obtain an accurate evaluation of the intensity of collision injuries. This study explores different machine-learning techniques to evaluate their effectiveness in predicting the severity of injuries in traffic crashes. The techniques include neural networks, support vector machines, decision trees, random forests, logistic regression, and gradient boosting. When it comes to road safety and the effects of crashes, having a comprehensive understanding of the degree of injuries sustained. This investigation of the literature analyzes the merits and disadvantages of various machine learning techniques to predict the severity of injuries sustained in a collision. Consequently, the generalized study showed that sophisticated models such as deep learning and ensemble analysis are more accurate in their predictions. Situations like this, on the other hand, are good for models such as logistic regression and decision trees because they are more easily interpretable and usable. As a result, difficulties related to feature selection, dealing with imbalanced data situations, and a lack of understanding of the offered models persist. The purpose is to raise awareness of how to use a range of multifaceted methods and combine knowledge about the domain with references to improve forecasts as well as develop more secure transportation systems.