A review of traffic accident perception models considering multiple influencing factors based on analysis framework optimization

The perception of traffic accidents is crucial for improving road safety. However, existing studies have limitations, including fragmented analysis of influencing factors, weak generalization of perception models, and the lack of a specific review framework in this field. This study proposes a traffic meta-analysis method to systematically review and quantify existing research on traffic accident perception, and ultimately identify which influencing factors and model structures can enhance the accuracy of traffic accident perception. Methodologically, traffic meta-analysis follows four key steps. First, it screens literature based on inclusion and exclusion criteria. Second, it scores the literature using literature quality assessment criteria. Third, it calculates the percentage improvement (enhancement rate) of the models proposed in the literature over the baseline in terms of accuracy. Finally, it evaluates the role of model structures and influencing factors in the literature by considering the weighted enhancement rate of literature scores, thereby comparing the performance of different perception models. This study constructs a dedicated analytical framework for traffic accident perception models and provides practical guidance for the application of artificial intelligence models in the field of traffic safety.