Research And Application Of Recommendation Algorithm Based On Latent Factor Model

The traditional latent factor model recommendation algorithm only uses the user's historical rating and partial bias items,and does not fully dig out the user's deeper potential information which leads to problems such as low recommendation efficiency and user satisfaction.In order to solve the problems of the traditional latent factor model recommendation algorithm,the item-based collaborative filtering recommendation algorithm is first studied,and the recommendation precision is improved by introducing the sequence of user browsing information;secondly,by supplementing the time factor and the sequence of user browsing information,etc.Optimize the traditional latent factor model to further improve the recommendation performance of the model;finally,the improved algorithm is fused,and the fused recommendation result is applied to the Web recommendation.The main research work is:(1)Aiming at the problems of insufficient user behavior sequence mining in traditional item-based collaborative filtering algorithms,a collaborative filtering recommendation algorithm based on item forward and reverse order is proposed.Consider the user's click process on items,and introduce the user's click sequence on the items in order to improve the problem of decreased recommendation precision due to the shift of user interest.(2)Aiming at the problems of user potential interest and sparse interactive information in traditional latent semantic models,a time decay function based on Newton's cooling law and a latent semantic moedl recommendation algorithm with scoring corrections are proposed.The implicit semantic model is optimized by modifying the order of user browsing items,adding bias items,and introducing a time decay function based on Newton's law of cooling.(3)The proposed collaborative filtering recommendation algorithm based on the forward and reverse order of items,the time decay function based on Newton's cooling law,and the latent factor model recommendation algorithm introduced with score correction are combined through equal weighted averaging method and dimensionalization processing.The fusion recommendation results are applied to the movie recommendation based on the Flask framework(MYSQL).The improved item-based collaborative filtering recommendation algorithm,the improved implicit semantic-based recommendation algorithm,and the improved fusion algorithm of the two algorithms are compared and tested on the MovieLens dataset.The results show that compared with the traditional recommendation algorithm,the improved item-based recommendation algorithm and the improved implicit semantic model recommendation algorithm,the proposed fusion algorithm improves the recommendation precision to a certain extent.