Recommendation method and system based on a generative adversarial network and double clustering

Abstract

The invention discloses a recommendation method and system based on a generative adversarial network and double clustering, and belongs to the technical field of computer application. The method comprises the following steps: firstly, reading the incomplete evaluation data set of a user-project establishing an incomplete evaluation data set of a project, then constructing a generative adversarialnetwork consisting of a generative network and a discriminant network, then predicting and filling missing evaluation values by utilizing the trained generative network, finally carrying out double clustering, and carrying out corresponding project group recommendation on different user groups according to sub-clusters obtained by the double clustering. According to the recommendation method and system, the trained generation network is used for filling the missing evaluation value, and the defects that a traditional method for filling the missing evaluation value such as the mean value (or the number of people) and linear interpolation is low in precision and large in error are overcome; And the filled complete evaluation data is clustered by using the double-clustering integration algorithm, so that the clustering result is more effective than that of a single double-clustering algorithm, the pertinence of a project group recommended to a specific user group is stronger, and the recommendation effect is improved.

Description

technical field [0001] The invention relates to the field of computer application technology, in particular to a recommendation method and system based on deep learning and bicluster analysis. Background technique [0002] With the development of information technology and Internet technology, recommending information that users may be interested in from massive information has become a research hotspot. Traditional recommendation methods can be divided into collaborative filtering recommendation methods, content-based recommendation methods and hybrid recommendation methods. As an important unsupervised data mining method, biclustering analysis technique obtains local structures and interesting sub-patterns hidden in the data by simultaneously clustering the rows and columns of the data matrix. In 2000, CHENG and CHURCH proposed the concept of biclustering, and gave the CC algorithm of biclustering, which uses a greedy iterative search algorithm to find biclusters, and rep...

AI Technical Summary

Problems solved by technology

In 2003, YANG and WANG proposed the FLOC algorithm. By randomizing the initial clustering, it corrected the shortcomings of the CC algorithm that cannot find overlapping double clusters, but the clustering results tend to fall into local optimum.

Most of the current recommendation methods and systems do not consider the situation of missing evaluation values ​​or use simple mean (mode) substitution methods, linear interpolation methods, regression prediction methods, etc., but these methods have problems of low precision and large errors

In addition, existing recommendation methods and systems are not accurate enough in recommending targeted items for different user groups.

Images