A User Behavior Clustering Algorithm Combines Association Rules and Multi-valued Discrete Features

Abstract

The purpose of user behavior clustering analysis is to analyze the features of core user groups. However, existing user behavior clustering algorithms cannot directly handle multi-dimensional and variable-length sequence, called multi-valued discrete features (MDF), which exists in user behavior data. This leads to lower utilization of data and a decrease in the accuracy of user behavior similarity calculations. Aim at the above problems, a clustering method that combines association rule mining and MDF is proposed. Firstly, association rules (AR) are introduced into the calculation process of the Jaccard distance (ARJD). Then, the center update method is modified based on the K-mode clustering algorithm. Finally, a user behavior clustering algorithm ARJDKM combining ARJD and the center updating method is proposed. Experimental results on real data show that the ARJDKM algorithm outperformed the other comparison algorithms in purity, entropy, silhouette coefficient (SC). Finally, the proposed algorithm improves the quality of user behavior clustering.