Research On Personalization Recommender Technologies Based On User's Interesting And Behavior Of Browsing

Content:With the rapid development of internet today, network information inflates day by day, facing this numerous information resource, the majority of internet users are increasingly felt themselves to be required information has become more difficult. Thus, the personalized service technology was born. The personalized service is realizes according to user's hypothesis. It's one target-oriented service mode. It carries on the collection, the reorganization and the classification through each kind of channel to the resources, provides to the user with the recommendation related information, satisfies the user the demand. The personalized service has broken the traditional passive form service pattern, can use each resources superiority fully, the initiative development take meets the user personalization need to serve as the goal.This article has conducted the research to the personalized recommendation system and the correlation technique. Firstly, the implicit information extraction and the direct information feedback has been analyzed, these two kind of commonly used internet user hobby extraction way respective merit and the shortcoming, and proposed in this foundation one kind uses half explicit half implicit hobby extraction way to obtain the internet user's interest information new internet user hobby extraction way. Secondly, this article has carried on the analysis to the common several kind of interest model, proposed the construction in the internet user interest retrogression foundation interest model expression. Finally, through carries on the analysis to the personalized recommendation system's design goals, in the internet user interest abates in the interest model foundation, uses the Pearson relevant algorithm computation user and the project similarity, and using the K most close neighbor sorting algorithm computation similar user "the neighbor area", then proposed based on the resources project personalized recommendation.In addition, but also to the Pearson relevant algorithm, this article carried on the contrast test based on the cosine similar algorithm and based on the adjustment cosine similar algorithm, the experimental result had indicated that must surpass other two algorithms based on the adjustment cosine similar algorithm. In the most close neighbor volume size to the personalized recommendation system quality's influence's test in discovered that when the most close neighbor volume surpasses 30 the most close neighbor volume's size to recommends system's quality influence not to be obvious. Based on user's coordination filter algorithm and based on the resources project's coordination filter algorithm's comparison test indicated that compares based on the resources project's coordination filter algorithm based on user's coordination filter algorithm has the superiority. When uses to promulgate between the internet user browsing behavior and the interest hobby relates half explicit half implicit hobby extraction experiment which explains the homepage detention long and the homepage tows is after a short while long and the internet user displays to page's hobby degree related, homepage triggering number and internet user hobby not obvious relations.