Analysis of Traffic Accidents Based on the Integration Model

Abstract

To enhance the safety of road traffic operations, this paper proposed a model based on stacking integrated learning utilizing American road traffic accident statistics. Initially, the process involved data cleaning, transformation, and normalization. Subsequently, various classification models were constructed, including logistic regression, k-nearest neighbors, gradient boosting, decision trees, AdaBoost, and extra trees models. Evaluation metrics such as accuracy, precision, recall, F1 score, and Hamming loss were employed. Upon analysis, the passive-aggressive classifier model exhibited superior comprehensive indices compared to other models. Based on the model’s output results, an in-depth examination of the factors influencing traffic accidents was conducted. Additionally, measures and suggestions aimed at reducing the incidence of severe traffic accidents were presented. These findings served as a valuable reference for mitigating the occurrence of traffic accidents.