Research On Interest Of Point Recommendation Algorithm Based On Social Network Check-in Behavior

With the development of smart phones and social networks,location-based social networks have been coming into and altering people's life gradually.In nowadays with fast development of data mining technology,recommendation system has become a critical direction in data mining field.Therefore,point of interest recommendation has become an important research hotspot.While the social network users can be helped to filter out the information that they are not interested in by the personalized point of interest recommendation algorithm.In such a way,the interest of users can be explored better to greatly enhance the using experience.This paper proposes a LDA-based weighted collaborative filtering algorithm based on the research of mainstream algorithms in recommendation system,and combining the score data with the analyses of comment text,through the analyses of theadvantages and disadvantages of traditional collaborative filtering algorithm.The algorithm mainly makes the study in the characteristics of user comment text,and the influence on each comment text from user ratings,so as to conduct the scoring prediction and recommendation jointly.Firstly,this paper summarizes several current main recommendation algorithms of the recommendation system,and analyzes the evaluation methods of recommendation systems.Secondly,this paper makes the study of the characteristics of user comment text,and integrates the LDA topic model into collaborative filtering algorithm,so as to alleviate the problem of data sparsityin traditional collaborative filtering algorithm.Thirdly,the influence of different comment text is adjusted through the way of setting weight based on the characteristics of the comment text in different scoring to achieve the purpose of improving prediction accuracy.Finally,the methods proposed in this paper are tested and compared with other algorithms by experiments on the data set provided by the Yelp website.The study of this paper is recommended jointly based on the user scoring and comment text data,and combined with the traditional collaborative filtering idea,which has the practical significancefor the recommendation system of modern social network platforms.By comparing with other algorithms,it is proved that the algorithm is more effective in predicting the accuracy of scoring.The accuracy in predicting scoring of this algorithm is proved to be more effective through the experiment comparison with other algorithms.