The present paper based on rough set theory is to analyze the reason of an e-commerce customers losing. The e-commerce is virtual, customers purchase behavior is random, and there is the 20/80 theory. The focus to the e-commerce customers losing predict is to bring enterprise 80percent profits or frequent buying clients, they will be the study samples. Therefore, we must first find out these clients from numerous customers, analyze their purchasing behavior, and it is one of the important links loss prediction. This process may be realized by customer behavior data clustering. We have analyzed the data in one e-commerce database, and according to a certain algorithm has classified these customers, one kind is superior customers, one kind is general customers, the rest is temporary customers. And a lot of questionnaire survey have been done to these kinds of customers, and then combining e-commerce expert opinions formed the customers data analysis and decision table, then the algorithm, which is the decision table blindly delete attribute reduction algorithm, is adopted to process the attributes reduction to the decision table. Then, we get the reduction table of the customers’ data analysis and decision. According to the reduction table, we summarize e-commerce customers’ loss decision rule. Through these decision-making rules, we can predict these losing customers, and take timely measures necessary to retain.
Attribute Reduction of Boolean Matrix in Neighborhood Rough Set Model
