Knowledge point analysis and prediction method based on graph knowledge tracking

By constructing a knowledge concept hypergraph and a learning path guidance graph, and combining two-way information dissemination and causal gating mechanisms, the problems of neglecting behavioral data and static graphs in existing knowledge tracing models are solved, enabling more accurate knowledge point analysis and personalized learning path prediction.

CN120975208APending Publication Date: 2025-11-18JIANGSU UNIV OF SCI & TECH
0 Cites 0 Cited by

Patent Information

Application Number
CN202510831103.0
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2025-06-20
Publication Date
2025-11-18

AI Technical Summary

Technical Problem

Existing knowledge tracking models ignore the rich behavioral data during students' answering process, and the use of static knowledge graphs cannot reflect the dynamic changes between knowledge points. Furthermore, they do not fully consider individual differences, resulting in low prediction accuracy.

Method used

We construct a knowledge concept hypergraph and a learning path guidance graph, connect multiple knowledge points through hyperedges, and combine bidirectional information propagation and causal gating mechanisms to dynamically capture students' learning paths and personalized behaviors. We use hypergraph convolutional networks and directed convolutional networks to embed nodes and integrate learning path and knowledge structure information.

Benefits of technology

It improves the accuracy and adaptability of knowledge point analysis and prediction, enhances the understanding and prediction ability of personalized learning paths, and improves the adaptability and generalization performance of the model in cross-knowledge point integrated application questions.

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

Abstract

The invention discloses a knowledge point analysis and prediction method based on graph knowledge tracking, and the method comprises the following steps: 1, constructing a knowledge concept hypergraph in which nodes are knowledge points and exercises, and the connection mode is hyperedge; 2, according to the answering sequence of the exercises of each student, a learning path guiding graph is constructed, and the expression performance of node embedding of the learning path guiding graph is enhanced through a two-way information propagation mechanism; 3, performing embedding fusion on nodes in the knowledge concept hypergraph and the learning path guide graph to obtain comprehensive embedding representation of exercises; step 4, updating the hidden representation of the knowledge state of the student step by step by using the comprehensive embedded representation through a gating loop unit; and 5, predicting the knowledge point mastering degree and the answering result of the student based on the hidden state. According to the method, behavior data and individual differences of the students are considered, so that the accuracy of student knowledge state prediction is remarkably improved, and personalized learning recommendation is effectively supported.
Need to check novelty before this filing date? Find Prior Art