/KnowledgeTracing

Customizable Deep Knowledge Tracing models and reproducible experimental scripts in one place.

Primary LanguageJupyter NotebookMIT LicenseMIT

Warning: This code is provided as-is for reproductivity. No updates and security patches are planned so far.

Knowledge Tracing

Customizable Deep Knowledge Tracing models and reproducible experimental scripts in one place.

Quick start

Install and run an experiment

$ poetry install
$ poetry shell
$ python main.py config/debug/debug.json

Alternatively, you can use pip instead of poetry: pip install -r requirements.txt.

Create your own config JSON file and you can start your experiment.

Datasets

We used prepared data by Zhang et al. https://github.com/jennyzhang0215/DKVMN

Dataset name KC size Link
ASSISTments Skill builder 2009-2010 110 https://sites.google.com/site/assistmentsdata/home/assistment-2009-2010-data
ASSISTments Skill builder 2015 100 https://sites.google.com/site/assistmentsdata/home/2015-assistments-skill-builder-data
ASSISTments Datamining competition 2017 https://sites.google.com/view/assistmentsdatamining/home?authuser=0
statics 1223
synthetic 50

Environments

requirements version
Python 3.7
CUDA 10.1
PyTorch 1.5.0

BibTex

Please cite our paper if you use the code.

Pre-training

Proceeding p.265 | Slide

To reproduce, read notebook/Results_ICCE2021.ipynb and run the same experiment.

# To appear
@article{panaccuracy,
  title={Prior knowledge on the dynamics of skill acquisition improves deep knowledge tracing},
  author={Qiushi Pan and Taro Tezuka},
  booktitle={Proceedings of the 29th International Conference on Computers in Education},
  year={2021}
}
Qiushi Pan and Taro Tezuka, Prior knowledge on the dynamics of skill acquisition improves deep knowledge tracing, Proceedings of the 29th International Conference on Computers in Education, November 2021 (ICCE2021). (to appear)

Knowledge State Vector Loss

To reproduce, read notebook/Results_ICCE2020.ipynb and run the same experiment.

@article{panaccuracy,
  title={Accuracy-aware Deep Knowledge Tracing with Knowledge State Vector Loss},
  author={Qiushi Pan and Taro Tezuka},
  booktitle={Proceedings of the 28th International Conference on Computers in Education (ICCE2020)},
  year={2020}
}
Qiushi Pan and Taro Tezuka, Accuracy-aware deep knowledge tracing with knowledge state vectors and an encoder-decoder architecture, Proceedings of the 28th International Conference on Computers in Education (ICCE2020), Online, November 23-27, 2020.

https://apsce.net/icce/icce2020/proceedings/paper_58.pdf

References