Research on Movie Rating Prediction Algorithms

In the era of data explosion of movie and video websites, “informahon overload” and “information labyrinth” have brought serious troubles, but movie recommender systems can efficiently solve such problems. In order to solve these problems, we propose the RF that uses the users' activity and rating to select suitable experimental data and proved this method can efficiently reduce RMSE and MAE of various recommendation algorithms. On this basis, we propose the MCBF-SVD, which is a movie rating prediction algorithm based on explicit data and implicit data from movie datasets to predict the future ratings of movies from users with a certain degree of activity. Firstly, the MCBF-SVD uses weighting factors to discuss the impact of movie categories on predicting future rating behavior of users, and also improves the filtering method based on movie categories. Finally, the MCBF is combined with SVD algorithm which has good performance in CF. Compared with several existing algorithms, our MCBF-SVD greatly enhanced the accuracy of rating prediction, and improved the scalability and efficiency of personalized recommender systems.