The Realization Of E-commerce Recommendation System Based On Web Mining Research

With the development of network technologies and the popularization of Internet applications, electronic commerce is now making a big progress. Increasing electronic commerce websites are building with intense competition to each other. Now, personalization service is one of the important ways to success for the electronic commerce websites that have more and more customers and commodities.For the purpose of improving the customer access and their loyalty and enhancing the capability of cross sale, the data mining and web mining technologies can be implemented in electronic commerce for obtaining useful knowledge from the data of business and customers accumulated by the electronic commerce websites to guide the personalization service.The definition and main implementation technologies of electronic commerce are introduced; the importance, development and status in quo of the electronic commerce personalization service are explained and summarized. That the web mining is befitting for the demand of the electronic commerce personalization service is pointed out. The definition, characteristics, process model and research situation of data mining are introduced. The process of web usage mining which is mostly used in electronic commerce personalization service is depicted. The technology, key methods and difficulties of web usage mining based on web log are discussed. In this dissertation, it is the web user visiting records that replaced the web log as the main resource of data, and combined with dynamic web page applications, the records can be collected actively and selectively. Thus the disadvantages of the web log mining can be avoided, and the requirements of electronic commerce personalization service can be satisfied. The structure of the records is illustrated, and relevant experiments demonstrate the data collection and user identification. The lab results prove that the new method have a small system response delay, reduce the amount of data waiting for mining, ease the job of data clean up, improve the precision and manipulation of user identification and make a well fundament for improving the implementation efficiency of data mining algorithm. Finally, a proposal on the record-based electronic commerce personalization service system's architecture, working process and web deployment are given by the paper.After briefly introducing the methods of path visiting analysis and association rules, which are widely used in web usage mining, the algorithm, process of user clustering and page clustering analysis based on the URL-UserID association matrix are traversed. That the trade data of users can be used to estimate the interest of users for the merchandise page is pointed out. The improved algorithm and computation process is detailed in the dissertation. Thereby, the right of element in the matrix can reflect how much users are interested in the web page of merchandise description.