Abstract: Using the data of customer′s behavior characteristics to predict the customer′s demand for service has significance to improve the quality of automotive aftersales service. This paper rough set and entropy theory are made use of to extract the characteristic attributes from a large number of customer behavior attributes, which have significant effect on the state of the automotive parts. The reasoning evidences are constituted by these characteristic attributes. The BPA（basic probability assignment）corresponding to every evidence is calculated by decision rules′ intensity. Meanwhile, the comprehensive evidence is calculated by using the D_S evidential theory to synthesize the BPA. The customers′ service requirements can be inferred by this method. The method is proved by a case that it can be used for car′s aftersales service knowledge reasoning.

Key words:  customer behavior, rough set, D_S evidential theory