Matrix transformation and decomposition commodity recommendation method based on point cut set graph segmentation

Abstract

The invention discloses a matrix transformation and decomposition commodity recommendation method based onpoint cut set graph segmentation. The method comprises the following steps: 1) acquiring basic data, and constructing an original scoring matrix; 2) converting the original scoring matrix into a bilateral block diagonal matrix by using a cut set graph segmentation algorithm based on community discovery, and splicing bilateral and diagonal blocks of the bilateral block diagonal matrix into a block diagonal matrix containing a plurality of sub-matrixes; 3) based on sub-matrixes in the spliced block diagonal matrix, executing a matrix decomposition algorithm to obtain a set of decomposition results; 4) predicting a blank score according to a set of decomposition results to obtain an approximate matrix of the original matrix; and 5) carrying out personalized commodity recommendation on the user according to the approximate matrix. According to the method, a matrix decomposition method based on a bilateral block diagonal matrix is used as an effective means for personalized recommendation, the problem of data sparseness in a recommendation system and the problem of low efficiency of a matrix decomposition algorithm are effectively relieved, and the time spent on recommendation is shortened while the prediction precision is improved.

Description

technical field [0001] The invention relates to the technical field of big data analysis and recommendation, in particular to a product recommendation method based on matrix transformation and decomposition of point-cut set graph segmentation. Background technique [0002] With the rapid development of the Internet, people's online activities have become more and more frequent. At the same time, more and more users have various behaviors on the Internet, which has generated huge and complex data. set. How to use these huge data sets to apply personalized recommendations to different scenarios is a hot issue that needs to be solved at present. Collaborative filtering algorithm makes full use of collective wisdom, by analyzing the behavior of a large number of users, mining some hidden patterns, recommendation algorithm based on collaborative filtering has been the most extensive and deepest research. Among them, the collaborative filtering based on the latent variable model...

AI Technical Summary

Problems solved by technology

[0005] 3) Interpretability of recommendation results. The decomposition results of matrix decomposition are based on hidden variables, but we cannot clearly determine what these hidden variables represent. It is difficult to give users an acceptable and convincing recommendation result

Images