Collaborative Filtering Recommendation Technology Based On User Behavior

Collaborative filtering is one of the most widely used recommendation technology.However,there are still some problems in this technology,such as the subjectivity of user rating behavior and data sparsity.Besides,collaborative filtering algorithm only analyses the historical rating data without considering the correlation of between users and tags,items and tags.In view of the above-mentioned problems,according to the analysis of user behavior,the collaborative filtering recommendation technology is studied.The main research contents are as follows:1.For the problem of the subjectivity of user rating behavior and data sparsity,a backpropagation neural network rating prediction algorithm based on cloud model is proposed.Firstly,the algorithm uses reverse cloud transformation to generate the cloud model for users and items.Secondly,the method of qualitative and quantitative conversion is utilized to generate multiple cloud rating prediction values for the users.Finally,aiming at the problem of data sparsity,the cloud layer composed of cloud rating prediction is added to the neural network to improve the rating prediction accuracy and then fill the missing value of rating matrix.Experimental results show that compared with traditional methods,this method can effectively improve the accuracy of rating prediction,alleviate inaccurate prediction caused by sparsity of user rating data,and improve recommendation performance.2.For the problem of inaccuracy of the user neighbors in the collaborative filtering algorithm,a bipartite graph material diffusion algorithm based on tag weight is proposed.Firstly,in order to describe user preferences more accurately,the “user-item” tag weight association matrix is constructed by using the tag rating as the weight calculation method.Secondly,according to the recommendation method based on graphs,the algorithm constructs a user-item bipartite graph with tag weights and uses the bipartite graph material diffusion algorithm to measure the similarity between users.Finally,a recommendation list is generated for the target user by combining with the user-based collaborative filtering recommendation algorithm.The experimental results show that the algorithm is superior to other comparison algorithms in recall,precision and F1 value,which effectively improves the recommendation quality.