Machine Learning Algorithms to Model Crash Severity and Collision(s)

Abstract

The fact that traffic accidents annually result in many lives, serious injuries, and economic losses makes them one of the most pressing problems the world is now dealing with. The issue of developing precise models to forecast the severity of traffic accidents is crucial for transportation systems. This research project develops models to choose a group of significant characteristics and to construct a system for classifying injury severity. Different machine learning techniques are used to generate these models. supervised machine learning methods. Currently, Random Forest, Support Vector Machine, Decision Tree, and K-Nearest Neighbor are the best methods for predicting the severity of injuries in traffic crashes. There is still a lot of opportunity to investigate other methods that can best serve this goal as not only the model's performance but also causality difficulties, unobserved heterogeneity, and temporal instability should be taken into account. Researchers may learn about the most recent strategies used in the study of injury severity modelling in this publication, as well as the ones that produced the greatest performance outcomes. Challenges and potential areas for future research are offered based on the examined papers.