GameDKT: Deep knowledge tracing in educational games

作者：

Highlights：

• This research developed a novel deep knowledge tracing approach for educational games.

• It models players skill level by integrating machine learning with domain knowledge.

• It predicts players performance in next game tasks using a CNN model with AUC of 0.913.

• The CNN model seems to be faster than sequential models in identifying local patterns.

• The CNN outperforms RNN and LSTM models in both one- and multi-step predictions.

摘要

•This research developed a novel deep knowledge tracing approach for educational games.•It models players skill level by integrating machine learning with domain knowledge.•It predicts players performance in next game tasks using a CNN model with AUC of 0.913.•The CNN model seems to be faster than sequential models in identifying local patterns.•The CNN outperforms RNN and LSTM models in both one- and multi-step predictions.

论文关键词：Learner model,Deep knowledge tracing,Educational game,Prediction of player performance,Deep learning

论文评审过程：Received 6 September 2021, Revised 27 December 2021, Accepted 8 February 2022, Available online 14 February 2022, Version of Record 17 February 2022.

论文官网地址：https://doi.org/10.1016/j.eswa.2022.116670