Subgroup Discovery Method Based on User Behavior Analysis

摘要

Subgroup discovery and analytics is an important tool for descriptive data mining. Most methods about subgroup discovery are based on the graph structure, ignoring the hidden mutual influence between nodes. In this paper, we consider the above issues, re-think the impact of implicit relationships between nodes, base on the network structure partition method, make the interaction behavior between user nodes reflected in the graph structure. A method of subgroup discovery based on behavior interaction (abbreviated as SDBI) is proposed. First, extract some features of data by using the method named "Initial classification of prominent data ", and then extract the remaining data in order according to the importance of the data, and construct the selected data node as a weighted complete digraph. Afterwards, the SDBI were used to divide subgroups. The results prove that the proposed method can effectively consider the problem of data non-uniformity and accurately divide the results.