Research on causation analysis and risk severity prediction methods for road traffic accidents

The safety of people's lives and property is greatly threatened by traffic accidents. However, there are numerous reasons for traffic accidents, and it can be challenging to identify the most important ones, which makes accident prevention challenging. Therefore, this study conducts relevant work using the UK Department for Transport's 2019 road traffic accident datasets. In order to determine the distribution characteristics between the accident information and the dimensional characteristics of the person, vehicle, road, environment, and accident form, this paper first processes the data from the road traffic accident datasets. It then uses multiple interpolations based on the chained random forest to fill in the missing values. Next, the study delves into the reasons behind vehicular accidents, enhancing the quantitative foundation of scenario clustering through the integration of Bayesian optimization-based random forest models, Cramer's V correlation test, K-Modes clustering, and frequency statistics, culminating in the identification of dangerous situations involving non-operating vehicles and passenger vehicles. Subsequently, the Apriori algorithm, based on the attribute values of the specified constraint items, is employed to conduct correlation studies across various dimensions, including individual, vehicle, road, environment, accident form, and temporal aspects, aiming to unearth a possible link between the severity of accidents and these dimensions. Ultimately, the model predicting accident risk levels employs the LightGBM algorithm, based on Bayesian optimization. The model underwent external validation and interpretive analysis using the 2022 UK traffic accident datasets, confirming its effective generalization capabilities.