GIKT

The code is related to the paper GIKT: A Graph-based Interaction Model for Knowledge Tracing published at ECML-PKDD 2020 Paper in arXiv.

Datasets

You can download the preprocessed dataset: download

If you want to process the datasets by yourself, you can download them by the following links:

ASSIST09:download

ASSIST2012: download

EdNet: download

You should create a folder named data for datasets. Besides, folders checkpoint and logs are also needed to create for saving model and log in our experiments.

Environment Requirement

python == 3.6.5

tensorflow == 1.12.0

numpy == 1.15.2

Examples to run the model

ASSIST09 dataset

• Command

python main.py --dataset assist09_3 --n_hop 3 --skill_neighbor_num 4 --question_neighbor_num 4 --hist_neighbor_num 3 --next_neighbor_num 4 --model hsei --lr 0.001 --att_bound 0.7 --sim_emb question_emb --dropout_keep_probs [0.8,0.8,1]

If you have more questions about our experiments, you can contact us. email: yyang0324@sjtu.edu.cn