Study On Recommendation Algorithm Based On Tags And Ratings Information

With the rapid growth of Internet users and the tremendous development of mobile communication terminals,how to obtain useful information from massive data has become an urgent problem to be solved.As an important means of information filtering,recommendation system(RS)is just the way to solve this problem.The core technology of RS is recommendation algorithm,while most recommendation algorithms are based on ratings and adopt traditional or improved similarity method to dig users' interests and make recommendations for them.While these methods are widely used,they do not make full use of other available information,such as labels,timing factors,and location.As a medium to connect users and items,tags can be combined with ratings and integrated into the calculation process of similarity,which can dig out more relevant information between users and items and further improve the recommendation quality.Therefore,based on the rating matrix,this paper considers the tags information and proposes two improved collaborative filtering(CF)recommendation algorithms based on the user and item similarity calculation methods.The main contents of this paper are as follows:(1)A CF recommendation algorithm based on tags weight is proposed.This algorithm use the tags information to calculate the similarity between users and weights the similarity with the scoring data.The algorithm makes good use of the relationship between users and tags information,and use tags information to distinguish the similarity between users.(2)A hybrid recommendation algorithm based on tags and ratings information is proposed.This algorithm introduces tags information and extends it in user and item dimensions,and then calculates the predicted value of target user's items by linear weighting.This algorithm comprehensively considers the relationship between scoring matrix and label information and effectively alleviates the sparsity of rating matrix.In this paper,compared with the traditional CF recommendation algorithm on MovieLens datasets,the experimental results and theoretical analysis of MAE,precision,recall and F1-measure verify that the proposed algorithms have better recommendation effect and improve the quality of RS effectively.