Using Fuzzy C-means to Discover Concept-drift Patterns for Membership Functions

People often change their minds at different times and at different places. It is important and valuable to indicate concept-drift patterns in unexpected ways for shopping behaviours for commercial applications. Research about concept drift has been growing in recent years. Many algorithms dealt with concept-drift information and detected new market trends. This paper proposes an approach based on fuzzy c-means (FCM) to mine the concept drift of fuzzy membership functions. The proposed algorithm is subdivided into two stages. In the first stage, individual fuzzy membership functions are generated from different training databases by the proposed FCM-based approach. Then, the proposed algorithm will mine the concept-drift patterns from the sets of fuzzy membership functions in the second stage. Experiments on simulated datasets were also conducted to show the effectiveness of the approach.