Heterogeneous network recommendation algorithm based on deep neural network

Abstract

The invention discloses a heterogeneous network recommendation algorithm based on a deep neural network. The heterogeneous network recommendation algorithm comprises the following steps: S1, representing vectors of global local information of users and articles; S2, automatically selecting meta-path types by utilizing a bolt genetic algorithm; S3, obtaining meta-path instances under the optimal Xmeta-path types; S4, obtaining an interaction vector based on the meta-path; S5, fusing global local information of the user and the article; S6, enhancing the vector representation of the user and the object by using a collaborative attention mechanism; S7, obtaining scores of the user and the article pair; S8, building loss function optimization parameters; and S9, repeating the steps 1-8, and when lu, i stably tends to a very small threshold epsilon (epsilon) 0), stopping training to obtain a heterogeneous network recommendation model based on the deep neural network. According to the invention, valuable meta-path types are automatically obtained by utilizing a genetic algorithm, so that the interference of human factors is reduced; and global and local information in the heterogeneousnetwork are mined through the node domain and the network structure information.

Description

technical field [0001] The present invention belongs to the field of machine learning, and specifically relates to a heterogeneous network recommendation algorithm based on a deep neural network. Background technique [0002] Different from homogeneous networks, heterogeneous information networks with different node and link types integrate complex information and contain rich semantics. Therefore, recommendation methods based on heterogeneous networks have proliferated in recent years. Although these methods improve the recommendation performance to a certain extent, there are still deficiencies. First of all, the meta-path types of most existing recommendation algorithms are usually defined artificially, and most of the meta-path types are judged based on prior information, which has certain interference. Furthermore, these algorithms seldom explicitly characterize meta-paths. At the same time, these algorithms also face the problem of how to extensively explore in hete...

AI Technical Summary

Problems solved by technology

Finally, these algorithms do not explicitly account for interactions between meta-paths and the user-item pairs involved

Images