A Dynamic Graph Sequence Recommender System Sensitive to User Interaction

Abstract

The present invention realizes a dynamic graph sequence recommendation system sensitive to user interaction through the method in the technical field of artificial intelligence. The system as a whole adopts a reinforcement learning framework. The data input is the user's rating data with time stamps on the products and the user's own attribute data. The output of the system is the recommended product sequence generated by consecutive rounds of recommendations. The recommendation results of each round are intelligent After observing the system environment modeled by the dynamic graph, the agent makes an optimal recommendation decision based on the state representation of the dynamic graph environment, the product representation, the user's real-time interest in the product, and user attribute information. The operation process of the system is divided into five modules in turn. It adopts the offline training method in reinforcement learning for training, uses the small batch gradient descent method to optimize parameters, and uses the graph neural network and self-attention mechanism to model the environment state, which can be based on real-time global The environment state generation recommendation strategy is recommended, which has strong real-time performance, high dynamics and scalability.

Description

technical field [0001] The invention relates to the field of artificial intelligence, in particular to a dynamic graph sequence recommendation system sensitive to user interaction. Background technique [0002] With the gradual deepening of social and economic informatization, problems such as information explosion and information overload are becoming more and more serious. Therefore, the way people obtain information is gradually changing from "people looking for information" to "information looking for people". As we all know, the recommendation system is an effective means to solve data overload. Accurate and effective recommendations can not only improve user experience and user stickiness, but also improve the efficiency of information transmission, which can directly or indirectly create more benefits. However, the user's interests and hobbies will change over time, and each interaction between the user and the recommendation system will be affected by its historical ...

AI Technical Summary

Problems solved by technology

However, on the one hand, these existing methods only consider the short-term benefits brought by single-step recommendation to users and the system in the recommendation process, which has great limitations; on the other hand, most of the existing methods are based on strong timing assumptions , adopting a sequence model or a static graph + sequence model to model the state of the environment, and this assumption is not applicable in all scenarios

Images