Research On Film Recommendation Algorithm Based On Mixed Strategy

With the rapid development of Internet,watching movies online has been attracting more and more attention as one of the ways to relax,and the competition between movie and video websites is becoming more and more intense.However,because there are too many movies on video sites,users need to spend a lot of time and energy looking for their favorite movies.The application of the recommended algorithm makes it easy for users to look for their favorite movies.Therefore,taking the movie recommendation algorithm as the research content,this paper proposes a Slope One algorithm based on dynamic threshold,and integrates it with collaborative filtering algorithm based on RBF neural network and collaborative filtering algorithm based on random forest,which makes the recommendation result more accurate.In summary,the main work is as follows:(1)The traditional Slope One algorithm is improved and applied to the field of movie recommendation.First,we use dynamic threshold to find the nearest neighbor set of target items.Then we use Slope One algorithm to predict the score of users' non scoring items,and finally get the recommended list,and verify the accuracy of the algorithm on MovieLens dataset.(2)A combination of RBF neural network and collaborative filtering is used in the field of movie recommendation.First of all,the initial user item matrix;then,calculate the similarity of the processed data,the similarity matrix based on similarity matrix to find the target item neighbor set;after using RBF neural network based on the neighbor set of data to predict the target user does not score to score;finally,according to the forecast score,M select the movie score higher recommended to the user,and verify the accuracy of the algorithm on the MovieLens data set.(3)The combination of random forest algorithm and collaborative filtering algorithm is applied to the field of film recommendation.First,calculate the similarity matrix according to the user's score;then,finding the user's nearest neighbor set,according to the neighbor set data,using the random forest algorithm to predict the user not scoring item score;finally,according to the prediction score level,before the M film selected higher scores are recommended to the user,and verify the accuracy of the algorithm in MovieLens data set.(4)Three kinds of recommendation algorithms are fused.The final recommendation results are based on dynamic threshold Slope One algorithm,collaborative filtering algorithm based on RBF neural network and random forest based collaborative filtering algorithm.In order to test the accuracy of the algorithm,the MovieLens real data set is used to detect the algorithm.The recommended results show that the fusion algorithm is better than a single algorithm.(5)A movie recommendation system is designed,including the acquisition of personalized information,system requirements,system architecture,function modules and processing flow of the system.