Abstract:Customer segmentation is one of the basic functions in customer relationship management. Most of the customer segmentation methods definitely classified the customer into different classes according to their attributes. But these definite classification methods can't fit the customer segmentation tasks well, for the uncertain and stochastic characteristics of customer behavior. The shortcoming of current definite customer segmentation methods consisted of a C-process and a P-process according to the uncertainty and the heterogeneity in data space of customers' behaviors. The proposed new model could meet the requirement in presenting the uncertainty of customers' behaviors by introduced the cloud model theory in to the P-process. A synthetic data experiment and a real data experiment indicated the effectiveness of this method