Mining Multi-Dimensional Weighted Association Rules in the Database of Dangerous Driving Scenes Based on a Hybrid Algorithm 2021-01-5021

The databases of traffic accidents contain many association rules related to the factors of dangerous driving scenes (DDS). In order to mine multi-dimensional weighted association rules in the databases according to the impact of each factor on traffic accidents, a hybrid data mining algorithm is proposed. It includes the Analytic Hierarchy Process (AHP) and Multi-Dimensional Weighted FP-Growth algorithm (MWFP-Growth). The AHP performs a hierarchical analysis of the DDS and establishes the judgment matrix between layers according to the method of pairwise comparison. After the eigenvectors related to the maximum eigenvalues of the judgment matrix are calculated, the weight of each factor is determined, and a multi-dimensional weighted dataset is established; The MWFP-Growth algorithm is used to mine association rules in the multi-dimensional weighted dataset. In order to obtain more shared prefixes, and store more items in the frequent pattern tree, the algorithm uses the support count of the item to sort transactions. The results show that the hybrid algorithm demonstrates improvement in the efficiency of execution and the accuracy of the calculation. In addition, according to the multi-dimensional weighted association rules mined by the MWFP-Growth algorithm, the frequent patterns of factors and the relationship between the factors and the types of traffic accidents are obtained.