Research On Movie Personalized Recommendation System Based On Matrix Decomposition And Xgboost

In the recent decades,with the continuous development of the Internet and the film industry,online movie resources have become more and more abundant,but at the same time,it has become more difficult for users to quickly find books that meet their needs from massive movies.In order to enable users to quickly find their favorite movies,collaborative filtering and personalized recommendation systems came into being.Commonly used recommendation algorithms include user-based or item-based collaborative filtering recommendation algorithms,matrix decomposition,etc.The above algorithms are all based on user item scoring matrices.Due to the observable factors of ratings,the user item rating matrix tends to be very sparse,resulting in poor recommendation results.In order to improve the effect of the traditional collaborative filtering recommendation algorithm,this paper proposes a movie personalized movie recommendation algorithm based on matrix factorization and Xgboost.The algorithm first uses SVD ++ to fill the user rating matrix,and then uses the completed user item rating matrix to cluster users and movies according to the rating vector,and then calculates the score between each user and the two sets of rating vectors for each type of user Similarity,construct the similarity feature of each user about each type of user,and do similar processing for movies.At the same time,the attributes of users and movies(such as user age,gender,movie category,etc.)are added to the model to construct a supervised model,and the supervised model is trained using the Xgboost integration algorithm to obtain a supervised model that will Score predictions.This paper applies the proposed MFXGB recommendation algorithm to the Movie Lens?100k dataset and compares it with the results produced by traditional collaborative filtering recommendation algorithms.The results show that the MFXGB recommendation effect is better than the traditional recommendation algorithm.This paper further verifies that filling the missing user rating matrix in the MFXGB recommendation algorithm and adding the user and movie's own attributes can significantly improve the algorithm.Finally,the recommendation algorithm is passed,and the top10 recommendation list is generated with good accuracy and recall.