Electronic Commerce Data Mining using Rough Set and Logistic Regression

Electronic commerce (E-commerce) has gradually been the mainstream of business. There may be some unpredictable but frequent problems such as delay in shipment, shipping errors caused by E-commerce participants’ low efficiency. There problems will have negative impact on the business of participants eventually. Correct evaluation of the efficiency of E-commerce is an important way to improve operations. This paper introduces the knowledge discovery theory of data mining-based on Rough Set Theory (RST) to deal with the vague and inaccurate information about the evaluation of supplier and mine the law knowledge that exists between input variables and adverse position. The output of RST is then used as the feature and is delivered to the Logistic Regression (LR) to rank the product of electronic commerce website. The proposed approach, termed as RST-LR, is composed of the procedure of attribute values discretization; filtration processing of minimum attributes sets; evaluation rule; calculating the ranking accuracy and the establishment of evaluation systems. We evaluated the proposed approach on a real world dataset, The experimental results show that it achievesa high accuracy, and the rule has met the requirements of application.