Recommendation method based on star map neural network

Abstract

The invention discloses a recommendation method based on a star map neural network. The method comprises the following steps: step A, generating a d-dimensional vector xi belonging to Rd through eacharticle xi in an embedded layer session, and enabling each session to be constructed into a star session graph; step B, inputting the embedded articles into a multi-layer star map neural network, andembedding the articles before and after the star map neural network by using a high-speed network; and step C, representing the session by combining general preferences and recent interests in the session; after obtaining the session representation, generating recommendations by calculating scores on all candidate items. Defects in the prior art can be overcome, and the recommendation effect is improved.

Description

technical field [0001] The invention belongs to the technical field of recommendation systems, in particular to a recommendation method based on a star map neural network. Background technique [0002] The recommendation system can help people obtain personalized information, and it is widely used in web search and e-commerce. Many existing recommendation methods utilize users' long-term historical interactions to obtain their preferences for recommendation, such as Collaborative Filtering (CF), Factorized Personalized Markov Chain (FPMC), and methods based on deep learning, etc. Accurately obtaining user preferences is challenging in situations where the user's long-term historical interactions are unavailable, such as new users. Session-based recommendation is to generate recommendations based only on ongoing sessions. Most of the current session-based recommendation methods use recurrent neural network (RNN) to focus on the temporal information between items, and the at...

AI Technical Summary

Problems solved by technology

However, they are prone to overfitting

Images