Answers are grounded in this paper and cite their sources.
Urban signalized intersections are high-incidence areas for traffic conflicts, particularly due to pedestrians' red-light running behaviors. The persistent occurrence of such behaviors significantly disrupts the order and efficiency of traffic flow intersections. Although previous studies have applied supervised learning approaches to detect these behaviors, they often fail to explore the interpretable causes. To bridge this gap, this paper proposes a supervised learning framework for pedestrians' red-light running behaviors, based on interpretable factors. A wealth of relevant video and questionnaire data from real-world signalized intersections was collected, based on which, potential causes of pedestrians' red-light running behaviors were extracted from four facets comprehensively: demographic-discipline, self-discipline, external-discipline, and environmental-discipline. Each instance was labeled with a binary value of 0 or 1 to indicate whether a pedestrian committed a red-light violation. A binary logistic regression analysis was performed further to assess the impact of individual factors and their combinations on pedestrians' red-light running behaviors. Utilizing these causes, five groups of supervised learning experiments were then performed to evaluate the classification performances of pedestrians' red-light running behaviors. The experimental results reveal two key insights: (1) the combination of most factors from self-discipline, external-discipline, and environmental-discipline facets is likely to induce pedestrians' red-light running behaviors; (2) all chosen classifiers achieved high detection accuracies, ranging from 85% to 96%, by leveraging the aforementioned interpretable causes. The optimal classifier identified in this research may show great potential for online detection of pedestrians' red-light running behaviors in new scenarios, offering a practical solution to enhance traffic safety and efficiency at urban signalized intersections.
No takes yet. Share an insight, caveat, or question.
Zhao et al. (2025) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: