Risk Assessment and Identification of Freight Vehicle Trip Chains

Although the number of trucks is low, their accident rate is high, and the consequences of accidents are severe. This paper is based on GPS data from 100 trucks, with each trip chain defined by a vehicle’s stay time greater than 20 minutes. The kinematic parameters for each trip chain are then extracted, and the entropy weight method is used to calculate the weights of various parameters. A random forest model is applied to select 11 key indicators, including speed and acceleration. The entropy weight-TOPSIS algorithm is used to assess the risk of each trip chain for the trucks. Different combinations of continuous and discontinuous trip chain scenarios are constructed. Finally, support vector machines (SVM) and decision tree methods are used for risk prediction under different trip chain combinations. The results show that the 11 selected key indicators provide an accuracy of 95.74% for describing the sample. In general, the SVM model shows better prediction accuracy than the decision tree under different trip chain combinations, though the decision tree results fluctuate significantly. As the penalty parameter in SVM and the minimum leaf node in the decision tree increase, the accuracy of the model gradually decreases.