Session recommendation method based on graph neural network GNN and multi-task learning

Abstract

The invention provides a session recommendation method based on a graph neural network (GNN) and multi-task learning, which comprises the following steps: acquiring click data of a user on an e-commerce website, and establishing a user session data set; constructing a user session directed graph according to the user session data; building a GNN-MultiTask-Learning neural network model, and training and outputing a user session representation; and calculating recommendation probabilities of all candidate items according to the output user session representation input scoring function, and carrying out personalized recommendation. According to the invention, for a session recommendation scene, the relationship between user click items and the influence of historical sessions on the current session are obtained, and implicit data is added through multi-task learning, so that user representation is more universal and transferable, the preference of a user to the items is estimated more accurately, the interest trend of the user is captured, and the click rate of the user to the project is improved.

Description

technical field [0001] The invention belongs to the technical field of session recommendation, and in particular relates to a session recommendation method based on graph neural network GNN and Multi-task Learning. Background technique [0002] In recent years, with the rapid development of Internet information volume, the data scale has also ushered in explosive growth. Massive data contains great value and potential. At the same time, people will also face the problem of information overload brought about by massive data. How to quickly obtain valuable information from a large number of data has become a key problem in the development of big data. For example, large-scale e-commerce platforms (Taobao, Amazon, etc.) often contain hundreds of millions of items, and it is difficult for users to find their favorite items among a large number of items. Recommendation algorithms can dig out the interests of users from massive data and give them Recommendation has become an imp...

AI Technical Summary

Problems solved by technology

Due to the different implementation methods of the two types of recommendation algorithms, each has its own unique advantages. The traditional neighbor recommendation algorithm has strong interpretability, simple implementation, fast model update, and is widely used in major e-commerce websites. Compared with Algorithms based on deep learning have poor performance; on the contrary, recommendation algorithms based on deep learning have excellent performance, among which the recurrent neural network model (RNN) has attracted much attention in session-based recommendation scenarios due to its excellent sequence modeling capabilities, and has achieved Good results, but there are still many problems

[0006] Problems in the current deep learning session recommendation model: (1) How to provide more accurate user interest prediction based on user session content and limited data information

Images