Personalized Recommendation Technology Based On The Study Of Social Labels

In recent years, the continuous development of the Internet and Web2.0has led tothe generation of a large number of information data. The phenomenon of informationoverload is becoming more and more serious. In this case, it becomes increasingly moredifficult to find the information people are interested from the large amount ofinformation. Therefore, the personalized recommendation comes into being. It is themost effective way to solve the information overload and it uses knowledge discoverytechnology to find target user’s interest.Social tag is one of the main applications of the Web2.0thinking. In Internetwebsite, users can freely mark their favorite information, which reflect the user’sinterests. If we introduce the tags of which we use to tagging the Internet informationinto the personalized recommendationto to form a personalized recommendation system,it must be much easier to analyze users’ preferences. Naturally, we can get betterrecommendations to satisfy users’ demands.Different from the former social tag based personalized recommendation systemscan only recommend a single type, recommending the item-tag joint recommendationcan recommend item and tag. And the tags user used to tag items reflects the user’spersonal preferences. So, it has great meaning to mining the topics of the tags which isused by user to tag items. Tags in item-tag joint recommendations can explain thereasons of recommended items, so the recommendation system is more interpretabilityand users’ acceptance will be higher. And recommending the joint recommendation foreach user makes full use of tagging data to produce high-quality recommended. Finally,we map the joint recommendation results into item space to get the final itemrecommendation result.In this paper, we use tag to improve collaborative filtering recommendationalgorithm and user-item-tag tripartite graphs diffusion recommendation algorithm to getitem-tag joint recommendation. In collaborative filtering algorithm, we use the diffusealgorithm to calculate the similarity between users. Then we consider binary relationsand ternary relations into collaborative filtering algorithm. Firstly, we get the joint item-tag recommendation results from item recommendations and tag recommendations.And then, we use a new user model which is represented by joint item-tag matrix to getthe joint recommendations. Finally, we combine the two results. In tripartite graphs, weconsider adding a recommended step. We study the relationship of item and tag to getitem-tag joint recommendation. And at the same time, we add time factor into the graphto achieve better recommendation.