Research On Recommendation Algorithm Based On Collaborative Filtering And Weighted Bipartite Graph

With the rapid growth of the data on the Internet,information overload is becoming a difficult problem,recommendation algorithm is one of the main ways to solve the problem.However the traditional collaborative filtering recommendation algorithm exists problems that the efficiency is not high enough,the accuracy is low and even memory overflow.In addition,with the increase of number of users the problem of data sparseness will be more serious and the results of the collaborative filtering recommendation algorithm will be less accurate.In order to solve the problems above and improve the efficiency and accuracy of the algorithm,this paper made some improvements,the specific work includes the following several aspects.Firstly,the method of data preprocessing was improved,the use of time decay curve can more fit the fact that the interest of users changes with time,which is more close to the real interest;Using the method of gaussian normalization to preprocess the rating data can eliminate the problem of the score standard is not unified caused by personal factors and the data after preprocessing become more specifications and standards.Secondly,using the improved hierarchical clustering algorithm and the K-means algorithm to divide the whole users into lots of similar user classes.The method redefined the clustering distance,which considered the influence of the common categories on the similarity of users and gave attention to the influence of the public items and the coincidence of user ratings.The algorithm n improved not only made the clustering results more in line with the real situation,but also laid a foundation for the community found of user interest.Finally,constructing a new model for the subclass of users and items on the basis of clustering and predicting the ratings by the concept of Bipartite Graph.This paper did some improvement based on the bipartite graph and put forward a new algorithm called weighted graph algorithm,this algorithm significantly reduced the complexity of matrix computation,improved the precision of prediction and improved the efficiency of recommendation,which provided a theoretical basis for future real-time recommendation.Through the three improvements above,this paper raised a new recommendation algorithm based on collaborative filtering and weighted bipartite graph.This algorithm can deal with the shortages of the traditional algorithm,alleviated the problems such as the data sparseness,the low precision of recommendation,the memory overflow and so on.At the same time,the efficiency and precision of the algorithm are also be improved.At last,did the experiments to verify the algorithm using the data sets of Movielens,the experimental results showed that the improved algorithm got good results in MAE,RMSE,accuracy of the recommendation and efficiency of the algorithm.