Raw Data.

To improve traffic safety in mixed traffic involving human-driven and autonomous vehicles, this study explored safety risk factors from multiple perspectives. Based on crash reports involving autonomous vehicles (AVs) in the California, United States, the XGBoost algorithm and Shapley additive explanations (SHAP) analysis were used to investigate the factors affecting accident severity. Association rule mining was employed to analyze the factors contributing to emergency braking events, based on field data from driverless taxi operations in China. Additionally, using data collected from questionnaires, the risk perception factors of different traffic participants were examined using the average degree of aggressiveness method. The results of three aspects analysis revealed that risk factors associated with mixed traffic were concentrated in areas such as weekdays, road sections, multiple lanes, roads with central medians, lack of control, and adverse environments. Finally, some safety improvement suggestions are recommended.