Bipartite Network Algorithm With Living Experience

With the advent of large data age,the information overload problem becomes more serious.Recommendation system plays an increasingly important role in life as the key technology to solve the information overload problem.This paper mainly merges the life experience into the bipartite graph recommendation algorithm to improve the algorithm performance and to alleviate the cold start problem.The main works are as follows:First,when we get shopping,the users could select goods,the goods also could attract users.The recommendation algorithm often overlooks the attractiveness of the goods to the users,but focuses on the user's choice.In response to this problem,we propose a bipartite network recommendation algorithm based on the shopping experience,which combines the interaction between the users and objects into the bipartite network algorithm and applies the final interaction to determine the recommended result.Second,feedback is a common phenomenon in daily life,it could enhance or stabilize performance.But in the bipartite network recommendation algorithms,the structure is obvious and there is little feedback between nodes.Aiming at this problem,we propose a equal symmetric bipartite network recommendation model which based on node feedback.In the model,all the nodes of the bipartite graph are regarded equally,and the feedback is established among the nodes of each part of the bipartite graph and between the user-object sides to enhance the recommended performance.Again,we compare the noise points in the image with the new users cold start problem in the recommended algorithm.We treat the bipartite network selection relationship as a digital image consisting of 0 and 1,and the new user is treated as the noise points.We rely on the local map features of the bipartite network,and use the digital image noise removal technology to connect the new users and the original data set.By controlling the proportion of joined new users,we make the bipartite network recommended algorithm experiments in different proportion conditions and the observe the algorithm performance.Accordingly,We proposed an algorithm to alleviate the new users cold start problem.Finally,in order to measure the proposed algorithms better,we selected four evaluation indexes,such as accuracy,recall,diversity and novelty.We selected Movielens,Netflix and Book Crossing three data sets.For the bipartite network recommendation algorithm based on the shopping experience,we introduce four representative bipartite network recommendation algorithms as the contrast experiments.For the equal symmetric bipartite network recommendation model based on the nodes feedback,we establish the equal symmetry model algorithm of three different bench algorithms to verify the rationality of establishing nodes feedback.In order to ease the new users cold start problem,we introduced two different template algorithms and made experiments under different conditions of adding new users.By integrating the life experience into the bipartite graph recommendation algorithm,we have done a good job of improving the performance of the bipartite network recommendation algorithms and alleviating the new user cold start problem well.These results also highlight the research value of this paper.