Research On Personalized Recommendation System Based On Bipartite Network Structure

Networking is the most typical social feature of the information society.The availability of information services in the online society has been greatly enhanced,and information infrastructure with high-speed,ubiquitous,cheap,and easy-to-use information has become widespread.A large number of information products and information services are pouring into the market.People can enjoy information services to the maximum extent and personnel can create,obtain,use,and share information.In such an information society,information is being produced and exchanged all the time.Although rich in information and broadening people's horizons,it also brought trouble to a certain extent.The most prominent problem is that people can't quickly obtain the part of information they want in massive amounts of information.Instead,their limited energy is consumed unconsciously by a large amount of unrelated information.Therefore,it is particularly important to study user-oriented personalized recommendations.In recent years,the research of personalized recommendation system has always been the focus of scholars in the field of big data.The personalized recommendation method based on bipartite graphs has attracted wide attention as a recommendation method with high accuracy and no restriction on project types.Firstly,this paper deeply studies the traditional recommendation algorithm and the bipartite graph recommendation algorithm.Secondly,it presents a comprehensive visualization of the classic MovieLens viewing dataset.For the problems of the weighted bipartite graph algorithm,the weight coefficients and similarity in the weighted bipartite graph resource allocation process in the bipartite graph structure are improved.Based on this,a fusion random forest algorithm is proposed and a weighted bipartite graph recommendation algorithm based on random forest correction is proposed.After the algorithm is improved and merged,the accuracy of the recommendation is improved,and the problem that only the relationship between the user and the project is considered in the algorithm based on the bipartite graph network structure and the influence of the interest preference is ignored,thereby enhancing the interpretability of the recommendation;The second is to develop a bipartite graph recommendation algorithm that introduces a trust network.Using the trust matrix and similarity matrix,the accuracy of recommending in the case of data sparsity is improved.Finally,according to the evaluation indexes of the recommendation system,the above algorithms were respectively verified by experiments,and the range and the best value of the key parameters were discussed.The final development completed a personalized recommendation system based on bipartite graphs.