Establishing and refinement of customer profile using transactional data

Customer profiling is becoming increasing important in modern retail industry. In this paper, we propose an approach which can either build customer profile from scratch or refine existing customer profiles. In our method, all the products in sale are attached with pre-defined labels. We build the customer profile according to the corresponding transactional data. The advantage of this approach can be summarized as follows: (1) The transactional data can provide better understanding of the characteristic of the customers; (2) The customer profile can be updated adaptively and reflect the newest behavior of the customer. Experimental results indicate the advantage of our methods.