Graph convolutional neural network session recommendation method based on structure enhancement

Abstract

The invention relates to the technical field of session recommendation, in particular to a graph convolutional neural network session recommendation method based on structure enhancement, and the method comprises the steps: firstly obtaining a session representation of a target session text; then generating a corresponding session graph based on the session representation, and recognizing noise items in the target session text through the session graph; resetting the attention weight of the noise item in combination with an attention mechanism so as to eliminate the influence of the noise item; and finally, calculating final prediction probability distribution, and performing project recommendation based on the final prediction probability distribution. According to the graph convolutional neural network session recommendation method based on structure enhancement, the noise item can be recognized and the influence of the noise item can be solved, so that the accuracy of session item recommendation can be improved.

Description

technical field [0001] The invention relates to the technical field of Internet big data, in particular to a graph convolutional neural network session recommendation method based on structure enhancement. Background technique [0002] Session-based recommendation is a recommendation mode for anonymous users or users who have not logged in. It plays an important role in today's major e-commerce platforms (Taobao, JD.com, etc.) or streaming media platforms (Douyin, YouTube, etc.) . In actual scenarios, sometimes only short-term historical interactions of users can be obtained, such as new users or users who have not logged in. At this time, the performance of recommendation algorithms that rely on long-term historical interactions of users will be limited in session recommendation, such as methods based on collaborative filtering or Markov chains. Therefore, session-based recommendation has become a research hotspot, and its goal is to recommend the next item (or commodity)...

AI Technical Summary

Problems solved by technology

However, the graph structure in the existing GNN (Graph Convolutional Network) is constructed in a static manner. Once the user's preference shifts, it is easy to bring noise items (information) to the graph structure, resulting in an inaccurate graph structure. At the same time, , the item-transition relationship usually contains noisy items caused by accidental or erroneous user clicks

However, the traditional attention mechanism will assign attention weights (importance coefficients) to each item in the conversation text, that is, noise items will also be assigned corresponding attention weights, which is easy to introduce noise information into the conversation representation, Causes the model to learn an inaccurate representation of user preferences, resulting in low recommendation accuracy

Images