Since customers' behaviors may change over time in real applications, algorithms that can be utilized to mine these drift patterns are needed. In this paper, we propose a GA-based approach for mining fuzzy concept-drift patterns. It consists of two phases. The first phase mines membership functions and the second one finds fuzzy concept-drift patterns. In the first phase, appropriate membership functions for items are derived by GA with a designed fitness function. Then, the derived membership functions are utilized to mine fuzzy concept-drift patterns in the second phase. Experiments on simulated datasets are also made to show the effectiveness of the proposed approach.