Abstract: Using correlation analysis, a flexible evaluation index system for intersection safety under different conditions was established. Based on the data envelopment theory and the model of C²R, a super-efficiency DEA model was established, and by combining the improved K means clustering algorithm (FKM) which had the initial cluster center, an intersection evaluation method was constructed. Taking the 10 intersections in the central district of Xining City as an example for evaluation, the safety indexes and safety grades of each intersection were obtained. The results showed that this method could effectively evaluate the safety of urban intersections; By using super-efficient DEA index clustering, the bias clustering phenomenon was avoided and the existing intersection safety evaluation methods were enriched.

Key words: city traffic, safety evaluation, traffic conflict, super-efficiency DEA, improved FKM