Study And Implementation On The Personalized Recommendation Service Based On Web Usage Mining

With the fast development and the widely usage of the Internet, WWW(WordWide Web) has become the important way for people to acquire information as well asservice as it has accumulated large amount of information on various area and rich incontent. However, due to the characteristics of various in pattern, fast updated, mixedwith too much redundancy or valueless ones, the information on the Web make itdifficult for the Web users to fast capture the information or knowledge needed. Theidea that how to fast as well as effective fulfill the information need of the users so asto fast provide high quality information service is on the spot light nowadays. Andsuch idea is drawing such an increasing attention from the public that people have payenough interest on this issue.The traditional information service mainly relies on the search engines to searchthe Web information, then provide the users with the information filtered. But there isone point missing in this method that it does not take the diversity of the Web users.Each user has a different background, habit and goal of online. In this way, the generalinformation provided by the traditional method could hardly satisfy the different needof every user. In other words, we need a pertinence effective personalized informationservice to suit the taste of every user, based on the likelihood of the users. Focus onthis problem, the thesis do beneficial research and exploration on the PersonalizedTechnology based on the Web Usage Mining.Firstly, this thesis introduces the theory of Web usage mining and its scope ofapplication, summarily talked about the theory of personalized service theory and itsrequirements, and deeply analyzes the process of Web usage mining in thepersonalized service at the same time.Secondly, this thesis deeply research on the personalized recommendationalgorithm of association rules and cluster separately. In the association rule method,the thesis analyzes the weakness of the Apriori algorithm, coming up with animproved strategy of pruning optimization and transaction compression. Then itshows this improving method practices in the user frequent access path mining. In thecluster method, the thesis designs method on how to calculate the user page interestand show them, how to formulate user usage transaction set. The thesis comes up withan improved hierarchical clustering algorithm and practice in the user-basedtransaction set cluster analyze, showing a personal recommendation strategy connection with the user usage file set.Then, this thesis elaborates the thought of design on the Personalrecommendation system based on the Web usage mining, analyzing and designing themain component modules of the system. In addition, the thesis introduces the detailsof each module, including user identification, behavioral data collection, userpreference analysis and personalized information recommendation.Finally, through the re-development of the small and medium-sized enterpriseinformation resources platform management system, based on the SSH(Struts, Spring,Hibernate) frame makes out the personal recommendation system and research on theusage record of the platform’s user. The recommendation result fit in with the users’actual visit. In this way, it verifies the effective of recommendation algorithm and theutility of the personal recommendation system.