Research On E-Commerce Personalized Service Based On Click Stream Analysis

With the rapid development and popularization of Internet, e-commerce is booming at an incredible speed. Development of e-commerce brings tremendous impact. Enterprises can expand their business with e-commerce platform to promote their products and increase profitability; users can stay at home situation, as long as the click of a mouse, they can buy the products and services they need. As enterprises increasingly competitive, it is generally in pursuit of differentiation, e-commerce began to develop in the direction of personalized service.But through the research on personalized service tools, we find that the defects of the current personalized service tool can not provide timely and accurate information, and can not take the initiative to track changes in user interests, needs the participation of the user. To solve these problems, this paper introduces the click-stream technology to change the current existing defects.This paper introduces the research on personalized service the domestic and international, find that the existed defects of personalized service, and then describe the theory of click stream and its advantage on personalized services. Click-stream data is a web log file in the main, this paper mainly on the web log mining and found that the user preferences and buying habitsFurthermore, this article focuses on fuzzy C means algorithm (FCM) and knowledge of the basic concepts, describes the basic principles and steps of the fuzzy C means clustering algorithm. After reviewed on Fuzzy clustering, this paper proposes SWDFCM algorithms, including FCM algorithm initialization, the distance function .According to FCM algorithm divided data equally and vulnerable to noise point, this paper use the sample point density as weight, according to the distribution of web data itself, avoid the defects of FCM algorithm and improve performance of the anti-noise .Finally, the experiment analysis is conducted to verify the feasibility and effectiveness of the improved algorithm. Cluster user access log with improved algorithm, mining user interest model to analyze the user's individual requirements. This article programmed with Matlab, through multiple simulations verify that the performance of the improved algorithm is stable and anti-jamming in a noisy environment. So it is more suitable for web data characteristics.