In a competitive market like mail order market, it is crucial for companies in the market to identify key factors that affect customer retention, and to make action plans based on the factors and the corporate strategy. Then, it is needed for data-mining process from customer's historical purchasing data, to find proper customer segment, to predict change of the specific customer segment, and to clarify the prediction factors.
In our paper, we propose an ensemble algorithm which combine logistic regression model, decision tree model, K-nearest neighbor algorithm. From some computational experiments for benchmark instances and real data, we show the robustness and usefulness of our method.