Research on the identification and comparative analysis of typical traffic accident scenarios for urban conventional buses

To identify typical traffic accident scenarios for urban conventional buses, this study analyzed 1,791 bus accidents that occurred over a two-year period in a certain city. Nine variables, including accident time, road type, number of lanes, and other relevant factors, were selected as features. Four unsupervised algorithms, including K-means and K-medoids, were used for analysis. The results indicate that the K-means algorithm performed best in terms of both sum of squared errors and average silhouette score, successfully identifying typical bus accident scenarios. Based on the clustering results, bar charts and network topology diagrams were created and compared with accident scenarios involving cars and car-powered two-wheeler accidents to further explore the characteristics of bus accidents. The study found that urban conventional bus accidents can be categorized into seven types, primarily occurring during the day, and are most frequent on dual-direction roads with 5 to 8 lanes. The side of the bus is the main collision point, with accidents more common when traveling straight. Side and rear collisions of buses at roundabouts are also typical accidents. In contrast, car accident characteristics differ significantly from those of buses. Car accidents often involve colliding with straight-moving powered two-wheelers, with collisions occurring more frequently at intersections, and frontal collisions being more common.