Personalization Web Pages for Site Users, Utilizing Users’Interests and Sequential Patterns Discovery

With the rapid growth of information on the Web and increase ofusers who are daily visiting the web sites, presenting informationproportionate to requirements of users who are visiting a specialwebsite so that they could find their desired information would beessential. Therefore, analyzing browsing behavior of bew usersand modeling this behavior has particular importance. The aim ofrecommender systems is guiding users to find their favoriteresources and meet their needs, using the information obtainedfrom the previous users’ interactions. In this paper, to predict theweb pages with high precision, a hybrid algorithm of clusteringtechnique, All-K th-Order Markov model, and neural network arepresented. For this purpose, in order to model users’ movementbehavior, after clustering those with the same interests, thesequential patterns are extracted on users’ sessions of each clusterusing all-4th-order Markov model. Next, in the step of pagesrecommendation to a current user, which is performed in anonline state, first, a current user session is assigned to a clusterusing neural network. Then Markov model created on the clusterwhich has the nearest match to the current session, is applied anda sequence of pages, which the users are interested to view, isincluded in the list of recommendation. The implementationresults demonstrate that the proposed algorithm has higherprecision and recall comparing to other recommender systems.

کلیدواژه ها:

Personalization Web Pages ، Clustering ، NeuralNetwork ، Markov Models