Application Research Of Personalized Recommendation Service Based On Web Clustering

With the increasingly fierce competition of the Internet market,the users has become the grandest fortune of the Websites,these Internet users urgently desire to obtain the most accurate information with the simplest and quickest method. They hope the Website can recommend the content which they feel interest about, but have not browsed according to the Characteristics of internet users.The need of recommendation will effect the method of Information Service.Firstly,it will force Websites to change the old method that provides all users with unified interface and the same contents.Secondly,it will make the method that focuses on SCD(Website centered design) be replaced with that focuses on UCD(user centered design),that is to say,the Websites should not only have information and service which aim at all users' common interest,but also be able to organize and adjust the information and service automatically according to every user's personalized characteristic and interest.Therefore,the technology of personalized recommendation service that establish corresponding measures to develop commodity service according to the users' different demands is quickly used by more and more enterprises.So the personalized recommendation technology has become an urgent and important research subject.What is more, the technology will have higher academic value and application prospect.Firstly,this paper summarized the present research status at home and abroad of Web clustering technical on personalized recommendation and analyzed the advantages and disadvantages of these algorithms.Secondly,we got the best cluster number and the corresponding cluster centers automatically by competitive agglomeration which can overcome the drawbacks of K-means algorithm.Thirdly,we put forward a fuzzy conceptual clustering algorithm F-CobWeb combining the fuzzy concept lattice and Low Bias Fuzzy Probability Utility,this algorithm can overcome the drawbacks of CobWeb algorithm.The research is carried out by combing theory analysis and simulation experiments.The main content in the paper is as follows: 1.The research status of Web clustering technical on personalized recommendation was researched completely.Firstly,two common clustering algorithms and the detailed process of Web data preprocessing were introduced. Then,this paper expounds the methods of obtaining the website's static topology from web pages and dynamic link structure from Web log records.2.A user clustering algorithm CAKPS based on K-means algorithm was proposed.Firstly,a user visit matrix was established integrating of the three visit factors:access sequence,browsing time,click frequency.Secondly,a new distance method that captures the structure of a web site is defined to measure the similarity between two users.3.A page clustering algorithm F-CobWeb based on CobWeb and Fuzzy Concept Lattice was proposed.Firstly,a Page Fuzzy Concept Lattice visit matrix was established integrating of the Web log records and Website structure in fuzzy contexts.Secondly,concept hierarchy was constructed by the Low Bias Fuzzy Probability Utility.Then,the concept representation and relationship of ontology was automatically generated,and pages clustering sets was describe by the ontology.4.A personalized recommendation system PRWCL was designed and implemented based on Web clustering.The system is composed of offline module, including data preparation and extraction of clustering sets,and online personalized recommendation engine etc.