Score prediction method for constructing local matrix based on graph random walk

Abstract

The invention discloses a score prediction method for constructing a local matrix based on graph random walk, and belongs to the field of personalized recommendation. Depending on user-article scoringmatrix, a user-article bipartite graph is constructed, random walk is carried out on the bipartite graph, and A users with the maximum node convergence probability after walk and articles are selected to form A anchor points; and for each anchor point, a random walk algorithm with restart is adopted to obtain a correlation between each node and the anchor point so as to distribute each node intoa corresponding anchor point neighborhood. Each anchor point and the neighborhood thereof form a local matrix, and score prediction is carried out in each local matrix by using a matrix decompositionmethod. The prediction scores of the A local matrixes are averaged to obtain a final prediction result. According to the method, anchor points are selected and neighborhoods of the anchor points are constructed based on graph random walk, so that errors caused by a traditional distance calculation process are avoided; starting from nodes, the nodes are distributed to different anchor point neighborhoods, and complete coverage of a large matrix can be achieved.

Description

technical field [0001] The invention belongs to the field of personalized recommendation, and more specifically relates to a score prediction method for constructing a local matrix based on graph random walk. Background technique [0002] With the advent of the web2.0 era and the great improvement of network bandwidth, various social networking platforms have begun to appear, and fragmented information has begun to flood people's lives. In order to solve the problem of information overload, the personalized recommendation system is more and more showing its important value. For example, in the field of e-commerce, the recommendation system builds the user's interest model based on the user's historical behavior information, calculates the user's liking for items they have not purchased, and then recommends items that the user may like. [0003] In practical applications, the prediction of user preferences is usually carried out by collaborative filtering method. The basic i...

AI Technical Summary

Problems solved by technology

[0005] Aiming at the defects of the prior art, the purpose of the present invention is to solve the technical problem that the rating prediction result accuracy of the recommendation system in the prior art is limited

Images