Research On Collaborative Filtering Recommendation Algorithm Based On Item Implicit Feedback And Multi-factor Fusion

Research on recommendation systems has always been a hotspot in data mining and played an important role in solving information overload.Among them,collaborative filtering is a relatively mature recommendation technology,but the current recommendation algorithm still has problems such as data sparsity,cold start,and insufficient implicit information mining.This paper improves the collaborative filtering process by mining implicit information of the item and fusing multiple factors,thereby affecting the accuracy of recommendation and alleviating data sparsity.Firstly,this article describes common recommendation algorithms,introduces in detail the principles and advantages and disadvantages of matrix decomposition methods in collaborative filtering,analyzes the research status and related issues of item implicit feedback and collaborative filtering algorithm process,and proposes improvement strategies.Secondly,in the current matrix decomposition recommendation algorithm,the impact of explicit information of users and items on the score is generally studied.The mining of implicit information in the matrix is not yet mature,and most algorithms in the process of mining implicit information ignore the mining of attributes of the item itself.In order to better integrate the impact of item implicit information on user ratings,consider whether the item is in the same field and the item's word-of-mouth effect,and integrate it into the matrix decomposition model,and then learn the user and item characteristics by minimizing the objective function Matrix,as far as possible to achieve the predicted score close to the actual score.Thirdly,for the phenomenon of insufficient data information mining and low utilization rate in the process of recommendation algorithm based on collaborative filtering,introducing user overlapping rating items,mean rating and user conformity rating factors,establishing a multi-factor fusion model to optimize the collaborative filtering algorithm.Mainly through matrix decomposition to reduce the dimensionality of the data,fusion user rating information and trust information combined with specific similarity calculation,effectively calculate the similarity between users,in order to improve the accuracy of recommendations.Finally,experiments are designed on the classical data set based on the itemt implicit feedback model and the multi-factor fusion collaborative filtering algorithm model.Through experiments,the optimal parameters of the algorithm are found,and then the data are analyzed synthetically in combination with the comparative experiments.The experimental results show that the model constructed in this paper effectively improves the accuracy of the recommended results.