Analysis of the Harmfulness of Abnormal Riding Behaviors of Electric Bicycles Based on Improved Multiclass Logistic Regression Model

Abstract

To analyze the harmfulness of abnormal riding behaviors of electric bicycles in-depth, the research focuses on the 2022 electric bicycle accident data in a specific city in China. Based on an improved multiclass logistic regression model, the relationship between different abnormal riding behaviors and the severity of electric bicycle traffic accidents is explored. Firstly, the severity of accidents is categorized into three levels as the dependent variable, while driver attributes and various hazardous driving behaviors serve as independent variables to construct the multiclass logistic regression model. Secondly, the model is optimized by eliminating irrelevant independent variables and improving the link function. Finally, the harmfulness of abnormal riding behaviors of electric bicycles is analyzed based on the results of the regression model. The results indicate that eight factors significantly influence the dependent variable, with three factors, including driving under the influence of alcohol, being more likely to lead to fatal accidents, requiring focused attention for intervention and regulation.