A method for predicting academic research topic trends based on temporal graph attention network

By adopting a time-series graph attention network in academic research topic trend prediction, combining time-series convolution and multi-head graph attention module, the problem of existing methods ignoring the topic correlation characteristics is solved, and a higher precision trend prediction is achieved.

CN119294610BActive Publication Date: 2025-03-25NANJING UNIV OF INFORMATION SCI & TECH
2 Cites 0 Cited by

Patent Information

Application Number
CN202411803536.7
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-12-10
Publication Date
2025-03-25
Estimated Expiration
2044-12-10

AI Technical Summary

Technical Problem

The existing academic research topic trend prediction methods only consider the timing characteristics of the academic research topic itself, but ignore the correlation characteristics between the topics, resulting in low prediction accuracy.

Method used

The timing characteristics of academic research topics are extracted through the timing chart convolution module, and the multi-head graph attention module captures the correlation characteristics between topics and generates prediction results through the prediction output module.

Benefits of technology

It improves the accuracy of trend prediction of academic research topics, can capture complex correlation characteristics more accurately, and has a larger receptive field and higher prediction accuracy than traditional methods.

✦ Generated by Eureka AI based on patent content.
Patent Text Reader

Abstract

The present invention discloses a method for predicting academic research topic trends based on a time-series graph attention network, including: collecting metadata of multiple papers and preprocessing the metadata; constructing a data set with keywords as academic research topics from the preprocessed metadata; building a time-series graph attention network model, and training and optimizing model parameters on a training data set; inputting a test data set into the trained time-series graph attention network model to obtain the results of academic research topic trend prediction. The present invention combines a multi-head graph attention mechanism with time-series convolution to comprehensively extract the time-series features and correlation features of academic research topics, which can effectively overcome the limitation that traditional neural network academic research topic trend prediction methods cannot utilize academic research topic correlation features; the present invention can improve the accuracy of academic research topic trend prediction, and can be applied to academic trend analysis and paper recommendation modules of journal paper databases.
Need to check novelty before this filing date? Find Prior Art