Research On Recommendation System Based On Bipartite Graph Resource Allocation

The development of the Internet and the advancement of technology have brought great convenience to people's lives,but at the same time they have brought about the problem of information overload.The current means of solving information overload is mainly personalized recommendation system technology.Among them,the collaborative filtering recommendation technology has always been the focus of people's research since its inception.In order to improve the recommendation accuracy of the recommendation recommendation system,many scholars optimize the recommendation system by introducing new influencing factors or innovative recommendation models.In recent years,more and more scholars have optimized the collaborative filtering model by means of bipartite graph theory.On the basis of the bipartite graph theory and the collaborative filtering idea,this paper further adds the resource secondary allocation theory,constructs the object similarity matrix with higher accuracy,and further optimizes the collaborative filtering recommendation technology.The verification of the Movielens dataset shows that the improved recommendation model can significantly improve the recommendation accuracy.The main research results of this paper are as follows:(1)Constructed a low-dimensional,high-accuracy item-category model.Based on the singular value decomposition theory,the original scoring matrix is transformed into a scoring matrix of items in each dimension,and each dimension can be regarded as a potential classification of the item.The potential classification of the item is then optimized by the control variable method.By analyzing the optimized item-class matrix,the classification result is reasonable and can be used as the basis for constructing the item-category bipartite graph.(2)The item similarity matrix with higher accuracy is constructed,and the collaborative filtering algorithm is optimized.The constructed item-classification matrix is mapped into an item-category bipartite graph,and then the score of the item to the classification is converted into the weight of the side in the bipartite graph,thereby constructing a weighted bipartite graph.Based on the weighted bipartite graph,the item similarity matrix is constructed by theresource secondary allocation theory.Combining the item similarity matrix with the item-based collaborative filtering algorithm effectively optimizes the collaborative filtering algorithm.