E-commerce Personalized Recommendation Based On Web Mining Technology Research

With the growing popularity of Internet and the flourish development of e-commerce, Business web sites based on Internet, are facing more and more intense competitions. So,the development of business sites from the "site-centric" to "user-centric" becomes a necessity.The problem for e-commerce must be solved necessarily that how we can effectively orgnize and use a large number of complex e-commerce information,and extract interesting patterns to understand the customers' behaviors, so as to improve the sites' struture or provide customers with personalized services.Web as a large widely distributed global information service center, which contains a rich and dynamic hyperlinked informations and informations of interview and use, provides a wealth of resources for personalized services of e-commerce. The extraction of the useful information contained on Web needs to base on dataming, therefore, to use the personalized recommendation based on Web mining, is an effective way to resolve these issues.This thesis make a very comperhensive study for Web-data mining based on CRM.The work and innovation are mainly embodied in the following five areas:(1) It presentd Cross-style Web Mining approach for Personalized Recommendation, that introduces Web-content mining and Web-structure mining based on Web-usage mining personalized recommendation.Based on the improved definition of sparse matrix, we can use different recommended method in different situations of Web sites: In normal state, the Web-usage mining personalized recommendations can be used still;But when the user-project rating matrix is cold-start or sparse, we can provide users with personalized recommendations according to the similarity between the page contents and page links.In addition, Web content and Web structure mining results representation use the same expression way for sending to the recommendation engine.(2) It presentd an Improved PAM (IPAM) algorithm based on dichotomy and use it for the personalized recommendation of Web structure and Web content data.IPAM algorithm effectively solves problems of over iterations and high time complexity in the traditional PAM algorithm; In addition, it added the weight of Web structure and Web content in pages into the IPAM algorithm, from reality, make clustering more practical.(3) It designed and realized a simple offline personalized recommendation system based on Web mining.Experiments show that IPAM algorithm is more effective than PAM algorithm,and solves the problem of hightime complexity.The personalized recommendation based on cross-type method of Web mining, can effectively compensate for the deficiencies of traditional personalized recommendation way, and significantly improve the accuracy of the recommended.