Cross-domain recommendation method based on stacked auto-encoder

Abstract

The invention belongs to the technical field of cross-domain recommendation algorithms, and particularly relates to a cross-domain recommendation method based on a stacked auto-encoder. Aiming at theproblem of data sparsity existing in pure cross-domain recommendation, the invention provides the cross-domain recommendation method based on the stacked auto-encoder, which can improve the score prediction accuracy and the classification accuracy of recommendation. According to the invention, the two models of the cross-domain stacked auto-encoder based on the user and the cross-domain stacked auto-encoder based on the project are learned at the same time, the learning results are compared, the optimal recommendation result is selected, and therefore the score prediction accuracy and the classification accuracy of recommendation are improved. According to the invention, cross-domain information is introduced into the automatic encoder so as to understand deeper nonlinear network structures of users and commodities. According to the invention, the sparsity problem is effectively solved by expanding the target domain user vector and combining deep learning, and the method is superior toother models in the aspects of score prediction and Topn recommendation.

Description

technical field [0001] The invention belongs to the technical field of cross-domain recommendation algorithms, and in particular relates to a cross-domain recommendation method based on stacked autoencoders. Background technique [0002] With the rapid development of big data and Internet technology, the data generated by various applications in the network has exploded, leading to serious information overload problems. It has become very difficult for users to find information that meets their individual needs from a large number of resources (such as text, images, videos, and commodities). The recommendation system is one of the key technologies to solve the above problems. Different from search engines, recommendation systems study users' historical behavior data, make statistics and analyze their interests and preferences, so as to guide users to find their own information needs and realize personalized recommendations. At present, this technology is widely used in e-c...

AI Technical Summary

Problems solved by technology

[0005] From previous research work, we can see that the traditional single domain recommendation algorithm has great limitations, especially the cold start problem and data sparse problem.

Although cross-domain recommendation can effectively alleviate these two problems in theory, pure cross-domain recommendation still has the problem of data sparsity, so even if it is not obvious to introduce other data, most users will only interact with a small number of items

Images