A Personalized Recommendation Algorithm Based On Improved PageRank And Spectral Method

Internet brings convenience but also trouble,huge information network makes searching for effective information.The emergence of recommendation system alleviates the trouble caused by "difficult choice".Recommendation system can provide users with purchase suggestions,help users make better decisions and meet the individual needs of different users.Recommendation system research has produced a lot of results so far,but there are still some shortcomings.Aiming at the problem that the recommendation accuracy and the recommendation efficiency are not high in the present recommendation algorithm,this paper studies and improves the traditional PageRank algorithm based on the proposed algorithm,and proposes a recommendation algorithm based on PageRank and spectral method(Fusion PageRank and Spectral Methods).In the iterative process of PageRank algorithm,the desired number of candidate set nodes is added to control the number of iterations of the algorithm.After each iteration,a threshold is used to trim the results of each iteration,remove the nodes that have less effect on the results,control the number of nodes participating in the iteration,and obtain a candidate set.Then,using the idea of spectral method,the normalized adjacency matrix of candidate set is established,the eigenvalues and eigenvectors of adjacency matrix are calculated,the weight is calculated according to eigenvalues,and then the distance from target node is calculated according to eigenvectors.Based on this,the similarity between target user and candidate node is evaluated.The proposed algorithm is tested on MovieLens dataset and Douban movie dataset,and compared with the three recommended algorithms in accuracy,recall and recommendation efficiency.The experimental results verify that the proposed algorithm has some improvement in recommendation quality and recommendation efficiency.Based on the proposed FPSM algorithm,this paper develops a simple film recommendation system,which includes data acquisition,data conversion and three main functional modules.First,the system from the Watercress film site using web crawler technology to crawl the movie name,user name,user ratings and comment time and other film data as the recommended data,and then data conversion,the movie name and user name to renumber and put into local files,Finally,using the FPSM algorithm proposed in this paper as the recommended algorithm,the film is recommended according to the processed data and the personalized recommendation results are displayed in the interface.