Sequential social recommendation method and system based on door mechanism and storage medium

Abstract

The invention relates to a sequence social recommendation method and system based on a door mechanism and a storage medium, and the method comprises the steps: carrying out sequence division of original consumption data of a user and friends, and carrying out the initialization to obtain user sequence data and friend sequence data; obtaining current interests of the user and friends based on a GRU neural network of a selection gate mechanism; acquiring the short-term interests of the friends by network splicing based on a selection gate mechanism, intializing the commodity data of the friends to obtain the long-term interests of the friends, and splicing the short-term interests of the friends and the long-term interests of the friends to obtain the final interests of the friends; obtaining friend influence based on a neural network of graph attention, and splicing the friend influence with the current interest of the user to obtain final interest of the user; calculating probability distribution of different commodities, performing model training according to the probability distribution, and recommending commodity information to the user according to a training model; and the interest of the user can be learned more accurately through a selection gate mechanism, so that the recommendation performance of the recommendation system is further improved.

Description

technical field [0001] The present invention relates to the technical field of recommendation systems, in particular to a sequence social recommendation method, system and storage medium based on a gate mechanism. Background technique [0002] The recommendation system analyzes the user's behavior and finds the user's individual needs, so as to recommend some products to the corresponding users, helping users find the products they want but are difficult to find. There are various existing recommender system models, such as RNN-based sequential recommender systems: given a series of historical user-item interactions, RNN-based sequential recommender systems try to predict the following by modeling the sequential dependencies of the given interactions A possible interaction, in addition to the basic RNN, long short-term memory (LSTM) and gated recurrent unit (GRU)-based RNN has also been developed to capture long-term dependencies in sequences; another example is the combinat...

AI Technical Summary

Problems solved by technology

There are various existing recommender system models, such as RNN-based sequential recommender systems: given a series of historical user-item interactions, RNN-based sequential recommender systems try to predict the following by modeling the sequential dependencies of the given interactions A possible interaction, in addition to the basic RNN, long short-term memory (LSTM) and gated recurrent unit (GRU)-based RNN has also been developed to capture long-term dependencies in sequences; another example is the combination of residual sequences and social Recommendation system, this recommendation system not only utilizes the user's own interests but also combines the influence of friends to improve the performance of the recommendation system, but the existing technology is not accurate in learning the user's interest in the process of learning the user's interest to express the user's real interest

Images