Personalized Recommendation Research Based On User Interest Mining

In order to solve the problem that user interest extraction is not accurate and recommendation accuracy is not high in personalized recommendation and seriously affecting the personalized marketing and information resources access and management efficiency,we proposed a personalized recommendation research based on user interest mining,and take sinamicroblog as an example,we proposed a personalized recommendation method based on deep word embedding and limited attention mechanism.First,in view of the user in social media platforms that represented by sina-microblog will post some content that have some interference for user interest extract and the problem of the shallow level of user interest representation,we use a deep pre-training language model,and TF-IDF to mine out the deep semantic representation of the microblog and user label,and then based on the similarity between microblog and label to filter the microblog that interference interest extraction,based on this we used the clustering algorithm to mine user individual interest.Then according to sociology and social network research in the mechanism of limited attention,we used the social relationship between the target users and their followers to mine the social intensity between users and the social interest of the target users.And we use the relationship between the limited attention mechanism and social influence to construct the social influence coefficient,then use the it to control the amount of information that the followers' interests are integrated into the target user's interest.And use the social influence coefficient to merge the user interest and social interest,finally get the target user's integrated interest to recommend microblog.Finally,we carried out two experiments base on the collected data.The experimental results show that the personalized recommendation method based on the deep word embedding and the limited attention mechanism can solve the problem of user interest extraction is inaccurate effectively,improve the efficiency of personalized recommendation,provide effective guidance for the personalized marketing and the acquisition and management of information resources.