A Decision-Making Model for Predicting the Severity of Road Traffic Accidents Based on Ensemble Learning

Abstract

A road traffic accident is one of the most common and heinous tragedies that can occur anywhere in the world. Better safety and management of the roadways can only be achieved by investigating the causes of the occurrences. The materials under consideration have been compiled with the objective of addressing a number of themes associated with classifications of road traffic accidents. Nevertheless, the investigators’ model and information are not adequate in terms of efficiency or incidence to lessen the catastrophic loss. So this study investigates the possibility of using an ensemble approach to increase accuracy in predicting accident intensity and identifying critical elements. Our work utilizes voting ensemble learning techniques, as well as other underlying base models (Decision Trees, K-Nearest Neighbors, and Naive Bayes) for predicting traffic incidents. Different machine learning metrics were used to compare these models. Comparatively, the Ensemble method surpasses competing base classifiers by 89% accuracy, 89% precision, 89% recall, and 89% F1-scores. Furthermore, the provided informative model excels others on the ROC curve metric, demonstrating that for traffic safety administrators and authorized players, it’s a reliable and dependable method to make rational decisions.