Research On Personalized Recommendation Algorithm Based On Collaborative Filtering

With the rapid development of Internet and artificial intelligence in today's society,people are getting more and more information.But at the same time,with the problem of information overload,users can't find interesting content in a huge amount of information,so personalized recommendation technology came into being.The personalized recommendation system is an effective way to solve the "information overload" and has been applied to various fields,such as e-commerce,movies,and social.Collaborative filtering and filtering technology has become one of the most successful technologies in personalized recommendation because of its easy implementation and cross-domain advantages.However,the traditional collaborative filtering algorithm affects the accuracy of the recommendation due to problems such as sparse data,cold start,scalability,and poor real-time robustness.Based on the analysis of user behavior and related data,a series of improvement measures are proposed for the defects.The recommended accuracy of the collaborative filtering algorithm improved by experiments will be improved.The main work of this paper is as follows:In order to solve the problem of data sparsity in traditional collaborative filtering algorithm and improve the accuracy of recommendation algorithm,a collaborative filtering algorithm based on user similarity Slope One matrix is proposed.The algorithm first uses the basic cosine similarity to calculate the similarity between users and ranks them from high to low according to the similarity.The first n similarly similar users generate the user similarity matrix,and then use the Slope One algorithm to the neighbor users.Unrated items are predicted and used to fill the original useritem scoring matrix.Finally,the traditional project-based collaborative filtering algorithm is recommended.The proposed algorithm model alleviates the problem of data sparseness and is predicted in the score.In the process,considering the influence of user similarity on the scoring result,the recommendation effect is better than the general recommendation algorithm based on collaborative filtering.In the existing collaborative filtering recommendation algorithms,the impact of user attributes and item labels on recommendation quality is not fully taken into account.Therefore,the influence of user-based attribute similarity and user-label-based user similarity on recommendation results is comprehensively considered in this paper.User-attribute similarity and user-label similarity are combined into user similarity calculation.The final user similarity is the weighting of user-attribute similarity and user-label similarity.By adjusting the use of weighting coefficients,the sparseness of user similarity matrix is effectively alleviated,and the accuracy of user similarity is further improved,so that the recommendation results are more accurate.Experiments show that the two improved algorithms are better than the original ones in data sparsity and accuracy of algorithm recommendation.The recommendation algorithm in this paper has a positive effect on improving user experience,and has a certain reference significance for the research of recommendation algorithm in other directions.filling User attribute.