Knowledge graph–based safety risk evaluation method for hazardous behaviors of road transport vehicles

OBJECTIVE: This study aims to develop a knowledge graph (KG)-based framework to quantify and analyze the impact of hazardous driving behaviors on road transport safety. METHOD: A top-down approach was adopted to construct a multilayered KG incorporating seven categories of hazardous behavior factors (C1-C7). Multisource accident datasets were integrated to map the relationships among hazardous behavior factors, accident types, and accident causes. The Criteria Importance Through Intercriteria Correlation (CRITIC) method was applied to calculate the safety risk levels of various hazardous behaviors. Cosine similarity analysis was used to quantify correlations between hazardous behavioral factors and calculated risk metrics. Furthermore, KG-based path reasoning was used to trace causal chains linking hazardous behaviors to accidents. RESULTS: Dangerous driving (C5) and driver technical competency (C1) emerged as the two most influential risk factor categories, with correlation coefficients of 0.995 and 0.987, respectively. Rear-end collisions were identified as the most probable accident type caused by C5, with a conditional probability of 0.5. Fatigue and speeding were identified as the most common behavioral triggers. KG pathway analysis effectively traced risk propagation paths, highlighting key links in accident causation. CONCLUSIONS: This study integrates the multidimensional correlation analysis of knowledge graphs with the weighting advantages of the CRITIC method, explicitly expressing the causal chain of "hazardous behavior-accident type-accident cause" through graph structures to comprehensively analyze the behavioral mechanisms of traffic accidents.