Research On Personalized Recommendation System Based On Tag

With the rapid development of information technology and Internet, the amount ofinformation in the network is rising sharply. How to help users find valuable informationquickly in large amounts of data information, and how to make the data in the networkinformation attracts more users’ attention is the problem to be solved. Personalizedrecommendation system is an important tool to solve the problem.In recent years, more and more researchers pay attention to recommendationalgorithm-based on labels. However, the traditional recommendation algorithm which isbased on user-tag only judges whether the items have been chosen or not, while ignoringthe study to the project information of user behavior. And most of the algorithms put thosepopular hot commodities to the users, and there was no consideration on the impact of therecommendation results of the project, which ignores the measure of diversity, novelty andother importance. Aiming at the above problems, this paper is based on the analysis of theexisting recommendation algorithm based on user tag in the research, proposes one kind ofimprovement-personalized recommendation of tag-based algorithm. The main workincludes the following aspects:①Careful researches have been made on the relevant technology of recommendationalgorithm.Simply introduces and compares several common recommendation algorithmsand their advantages and disadvantages. And a detailed analysis of the recommendationalgorithm based on tags is made.②Secondly, consideringthe user tagdata affecting recommend novelty andinterpretability, using the tag constitutes a recommendation system labels, and labelsfordifferent user recommendation results were analyzed to get more reasonablerecommendation results.③In order to reduce the hot labels corresponding to hot items for a large weight,improve the novelty of recommended results, using the improvements of the label vector ofthe userfor its interest modeling, to reduce thetimes of label on hot items, increasing the novelty and accuracy of the algorithm.④Test the recommendation algorithm based on improved labelwith the data providedby Movie Lens. Analyses are made inaccuracy, diversity and novelty in three aspects. Theresults of experimental show that the improved algorithm has better performance in thesethree aspects, and verified the rationality andavailabilityof the algorithm.