Dynamic Integration of Preference and Knowledge Status for Knowledge Concept Recommendation (DISKRec)

This repository provides the official implementation for our paper "Dynamic Integration of Preference and Knowledge Status for Knowledge Concept Recommendation", published in Neurocomputing.

DISKRec is a dynamic knowledge concept recommendation model that jointly models learners' preference and knowledge status from learning and assessment behaviors to better capture their evolving learning motivations.

Fig. (a) The overall architecture of DISKRec, which consists of two modules: Status Disentanglement using Dual-DGNNs, Status Integration and Recommendation. (b) The steps in Dual-DGNNs take the initial graph status and a batch of interactions as input, and output the final graph status after three sub-modules: (1) Interaction Encoder, (2) Dynamic Status Updater with Dual-state Integration, (3) Neighbor Propagator.

Environments

• torch=2.1.1+cu118
• torch-scatter=2.1.2+pt21cu118
• torch_geometric=2.5.3

Code structure

DISKRec-anonymous
├── data
│   ├── model_input
│   │   ├── ednet
│   │   └── mooccubex
│   └── model_output
├── evaluation
│   └── metric.py
├── model
│   ├── combiner.py
│   ├── decayer.py
│   ├── dgnn.py
|   ├── diskrec.py
|   ├── edge_message.py
|   ├── graph.py
|   ├── iterative_updater.py
|   ├── neighbor_algorithm.py
│   └── rate_layer.py
├── util
│   ├── data_util.py
│   ├── early_stop.py
│   ├── global_config.py
│   └── temporal_interations.py
├── README.md
└── run_diskrec.py

Dataset

We provide the dataset to validate the effectiveness of our method.

Train and Test

Please unzip the dataset before running the code.

unzip data.zip

Then, run the model using the following command.

python run_diskrec.py

Citation

If you find our repo useful, please consider citing:

@article{liang2025DISKRec, 
  title={Dynamic integration of preference and knowledge status for knowledge concept recommendation},
  author={Liang, Qingqing and Lu, Xuesong and Wang, Chunyang and Qian, Weining and Zhou, Aoying},
  journal={Neurocomputing},
  pages={131786},
  year={2025},
  publisher={Elsevier}
}

For any questions or clarifications, please contact: qqliang.dase@stu.ecnu.edu.cn