Traffic Fatality Prediction Using Machine Learning Algorithms: Performance Analysis and Comparison Study

Abstract

Many nations are concerned about the high number of traffic fatalities, which has led to a search for efficient prevention methods. This study addresses road fatalities brought on by drunk driving incidents. The study uses machine learning algorithms to detect and comprehend patterns related to fatalities brought on by drunk driving. The study builds prediction models that can help with early detection and policy development using a number of variables, including crash features, demographic information, and historical trends. This models are formulated by various machine learning techniques. Supervised machine learning algorithms, such as Random Forests (RF), Decision Tree (DT), Naïve Bayes (NB), Logistic Regression (LR), Support Vector Machine (SVM) are implemented on traffic fatalities dataset. The findings of this study indicate that the RF model can be a promising tool for predicting of death by drunk driving. RF algorithm has shown better performance with 100% accuracy than DT with 95%, NB with 74%, LR with 94% and SVM with 67% accuracy.