基于SVR的城市道路刮擦事故评估模型

To effectively evaluate scratch accidents on urban intersection roads, a road scratch accident assessment method is developed based on the historical accident database and environmental parameters. Four environmental characteristic parameters; the location of the area where the accident occurred, weather conditions, road surface condition, and relationship between accident location and road were extracted by variance analysis and used as input for the evaluation model. The functional relationship between the four indexes and severity of the accident was expressed by the monthly traffic accident index, and the index was used as the output of the evaluation model. Particle swarm optimization, K-fold Cross Validation, and a genetic algorithm are used to calibrate the learning parameters of support vector regression (SVR), so that the model obtains the optimal parameters, and an urban road scratch accident assessment model based on SVR regression algorithm is established. The results show that the model is accurate and can be used for the risk assessment of urban road scraping accidents based on the above four indicators.