Graph Attention Neural Network Model With Behavior Features for Knowledge Tracking
The existing deep knowledge tracking models ignore the students’ behavior features and the high-order relationship between and questions with overlapping skills in the learning process. As a result, the models cannot learn the complete learning track of students and the dependence between...
Saved in:
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
IEEE
2023-01-01
|
| Series: | IEEE Access |
| Subjects: | |
| Online Access: | https://ieeexplore.ieee.org/document/10198404/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1849718721136820224 |
|---|---|
| author | Wei Zhang Sen Hu Kaiyuan Qu |
| author_facet | Wei Zhang Sen Hu Kaiyuan Qu |
| author_sort | Wei Zhang |
| collection | DOAJ |
| description | The existing deep knowledge tracking models ignore the students’ behavior features and the high-order relationship between and questions with overlapping skills in the learning process. As a result, the models cannot learn the complete learning track of students and the dependence between students’ historical answer records and questions, which affect the predictive performance of the model. In order to solve the above problems, a graph attention neural network model with behavior features for knowledge tracking (GAKT-BF) is proposed in this paper. Firstly, GAKT-BF designs a student learning behavior feature information extraction module, constructs a learning behavior feature matrix, incorporates students’ behavior features into the questions representation, and designs a new questions representation method. Then, GAKT-BF designs a question relationship extraction module and constructs a graph of correlations between questions and questions, uses graph attention neural network (GAT) to extract question vector representations, and finally predicts students’ next answers through long short-term memory (LSTM). Experiments on Assistments 2009, KDD CUP and Assistment-17 datasets show that GAKT-BF has significantly improved in both evaluation metrics AUC and ACC compared with existing advanced knowledge tracking models, and has better prediction results. |
| format | Article |
| id | doaj-art-57a4a3df4087426b836a8da262cda90b |
| institution | DOAJ |
| issn | 2169-3536 |
| language | English |
| publishDate | 2023-01-01 |
| publisher | IEEE |
| record_format | Article |
| series | IEEE Access |
| spelling | doaj-art-57a4a3df4087426b836a8da262cda90b2025-08-20T03:12:19ZengIEEEIEEE Access2169-35362023-01-0111883298833810.1109/ACCESS.2023.330070310198404Graph Attention Neural Network Model With Behavior Features for Knowledge TrackingWei Zhang0Sen Hu1https://orcid.org/0000-0003-1507-0215Kaiyuan Qu2Department of Artificial Intelligence Education, Central China Normal University, Wuhan, ChinaDepartment of Artificial Intelligence Education, Central China Normal University, Wuhan, ChinaDepartment of Artificial Intelligence Education, Central China Normal University, Wuhan, ChinaThe existing deep knowledge tracking models ignore the students’ behavior features and the high-order relationship between and questions with overlapping skills in the learning process. As a result, the models cannot learn the complete learning track of students and the dependence between students’ historical answer records and questions, which affect the predictive performance of the model. In order to solve the above problems, a graph attention neural network model with behavior features for knowledge tracking (GAKT-BF) is proposed in this paper. Firstly, GAKT-BF designs a student learning behavior feature information extraction module, constructs a learning behavior feature matrix, incorporates students’ behavior features into the questions representation, and designs a new questions representation method. Then, GAKT-BF designs a question relationship extraction module and constructs a graph of correlations between questions and questions, uses graph attention neural network (GAT) to extract question vector representations, and finally predicts students’ next answers through long short-term memory (LSTM). Experiments on Assistments 2009, KDD CUP and Assistment-17 datasets show that GAKT-BF has significantly improved in both evaluation metrics AUC and ACC compared with existing advanced knowledge tracking models, and has better prediction results.https://ieeexplore.ieee.org/document/10198404/Knowledge trackinggraph attention neural networkattention mechanismdeep learning |
| spellingShingle | Wei Zhang Sen Hu Kaiyuan Qu Graph Attention Neural Network Model With Behavior Features for Knowledge Tracking IEEE Access Knowledge tracking graph attention neural network attention mechanism deep learning |
| title | Graph Attention Neural Network Model With Behavior Features for Knowledge Tracking |
| title_full | Graph Attention Neural Network Model With Behavior Features for Knowledge Tracking |
| title_fullStr | Graph Attention Neural Network Model With Behavior Features for Knowledge Tracking |
| title_full_unstemmed | Graph Attention Neural Network Model With Behavior Features for Knowledge Tracking |
| title_short | Graph Attention Neural Network Model With Behavior Features for Knowledge Tracking |
| title_sort | graph attention neural network model with behavior features for knowledge tracking |
| topic | Knowledge tracking graph attention neural network attention mechanism deep learning |
| url | https://ieeexplore.ieee.org/document/10198404/ |
| work_keys_str_mv | AT weizhang graphattentionneuralnetworkmodelwithbehaviorfeaturesforknowledgetracking AT senhu graphattentionneuralnetworkmodelwithbehaviorfeaturesforknowledgetracking AT kaiyuanqu graphattentionneuralnetworkmodelwithbehaviorfeaturesforknowledgetracking |