Personalized Web Design Based On Web Log Mining

With the rapid development of Internet, people need to be morecomprehensive network services. Under the personalizing circumstance,thispaper builds up a model that user is interested in.This model gets user sinteresting vector by using his own information and other user s information thatthey left when they visited before,then filters the retrieval results to make themmeet user s needs. Although the traditional site meets the needs of people accessto information, but because of its universal nature, it can not satisfy the needs ofdifferent backgrounds, different purposes and in different periods in the visitor'sneeds.In today's highly competitive e-commerce environmet, the personalizedrecommendation has emerged as acritical application which is essential to a Website to retain visitors and turn casual browsers into potential customers.However,there are wide gaps between our country and other developed countiesabout the personalized recommendation,which is mainly caused by the scarcityof academic research.Under this background this dissertation applies the theoriesand methods of Web mining to the personalized recommendaton in e-commerceand utilizes the theory of comprehensive information and the model ofinformation moving process to system atically research the principle andmethods of the personalized recommendation in e-commerce.First of all, the problem of the Web user's interest analysis andrecommendation based on the syntactic method is discussed.The framework of the Web user's interest analysis and recommendation is described firstiy,and theprocess of discoverying Web transations is discussed in detail.Secondly,the problem of discoverying and matching recommendation rulesbased on Web text mining is discussed.The model of discoverying and thematching recommendation rules based on Web text mining is constructedfirstly,and the process of describing Web text based on Vector space Model isanalyzed,then the method of discoverying and matching recommendation rulesbased on the cluster of Web features items is discussed at length.Finally,the personalized recommendation method dased on Web semanticknowledge is discussed,which delongs to the semantic method of personalizedrecommendation in e-commerce.The framework of this method is proposedfirstly,and the construction process of Web domain ontology is illustrated by theeample of the bofei Website.Accoring to the Web domain ontology,aggressivesemantic Web usage profile and user active session are discovered then thepersonalized recommendation based on Web domain ontology is analyzeddetailedly based on semantic similarity measurement complex objects.This paper BoFei shopping system as a vector, a personalized Web sitemodel. Personalized recommendation is built on the Web based on the use ofexcavation. Mining is the use of Web server log files and customer transactiondata mining meaningful user access patterns and potential customer base, byadjusting the site provides links to individual users visit pages, in improving theefficiency of users visit at the same time increase user interest in the site visit, increase their market competitiveness.We describe the algorithm and processes of user access pattern, and presenta data preprocessing mode. First of all, we take several steps on web log to findusers' useful information, such as web site's URL, the time which the user stayon the web site, the times which the user visits the URL and etc. According tothese information, we can calculate the users' interests on the page he or she hadvisited. And the, we would recommend some pages which users haven't visit butmaybe favorite to users on the section of Favorite.We present the method of adjusting web page automatic in personalizedweb site. According to result of web usage mining, the program add dynamichyperlink automatic by inserting ASP code in web page to present different viewto unlike user. And the source program is given. The feasibility of adaptive websites has been tested by experiment.