Research On Fusion Matrix Decomposition Based On Trust And CF Recommendation Algorithm

With the rapid development of the contemporary Internet,e-commerce is rapidly emerging with the development of computer science,giving users convenience.Simultaneously,the matter of information overload is becoming more and more prominent in the process of intelligent evolution.The intelligent agent tool for information filtering-recommended system(RS)came into being.The recommendation system collects and learns user information,studies user preferences,and provides users with more accurate and comprehensive information recommendation services so that users can obtain the information they want at the lowest cost.A personalized video recommendation system that wins users and merchants by integrating and optimizing resources,and assists users to automatically,obtain videos that meet their interests and needs from the huge data information of the Internet,ensuring user experience and enhancing commercial interests.The maximization of the value of resources.The main work of this topic is:Firstly,an improved model based on Funk SVD is proposed for the sparse problem of scoring items in traditional recommendation methods.The stochastic gradient descent method is used to solve the minimum loss function SSE.Adding a regularization factor to improve the efficiency of the solution and establishing a preference prediction model for the unrated project,solving the data sparse aspect,and improving the recommendation accuracy;Secondly,in view of the impact that the trust relationship between users may have on the recommendation result,the strategy of introducing the trust relationship as the recommendation weight into the prediction score is adopted,and the accuracy of the algorithm in the interpretable recommendation is improved;Thirdly,for the traditional algorithm to rely too much on historical data to induce the cold start problem,the strategy of introducing the age and gender in the user feature into the user similarity calculation process is used to correct the similarity of the user to accurately locate the nearest neighbor set.Furthermore,the problem of user cold start is solved;in addition,the movie type in the project feature is introduced in the project similarity to calculate,and the similarity of the project is corrected to accurately locate the recent similar project set to solve the cold start problem of the project;Fourthly,a CF recommendation method based on trust degree fusion matrix decomposition and mining user project information is proposed based on the above strategy.The effects of global and local features on recommendation quality are considered comprehensively.Contrast experiments on the Movie lens data set show that the results show that The algorithm proposed in this paper is superior in recommending accurate accuracy.