华东师范大学学报(自然科学版) ›› 2022, Vol. 2022 ›› Issue (6): 68-78.doi: 10.3969/j.issn.1000-5641.2022.06.008

• 计算机科学 • 上一篇    下一篇

基于堆叠门控循环单元残差网络的知识追踪模型研究

黄彩蝶1, 王昕萍1, 陈良育1,*(), 刘勇2   

  1. 1. 华东师范大学 软件工程学院, 上海 200062
    2. 华东师范大学 基础教育与终身教育发展部, 上海 200062
  • 收稿日期:2021-08-10 出版日期:2022-11-25 发布日期:2022-11-22
  • 通讯作者: 陈良育 E-mail:lychen@sei.ecnu.edu.cn

Research on a knowledge tracking model based on the stacked gated recurrent unit residual network

Caidie HUANG1, Xinping WANG1, Liangyu CHEN1,*(), Yong LIU2   

  1. 1. Software Engineering Institute, East China Normal University, Shanghai 200062, China
    2. Basic Education and Lifelong Education Development Department, East China Normal University, Shanghai 200062, China
  • Received:2021-08-10 Online:2022-11-25 Published:2022-11-22
  • Contact: Liangyu CHEN E-mail:lychen@sei.ecnu.edu.cn

摘要:

知识追踪任务是根据学生历史做题记录和其他辅助信息追踪学生知识水平的变化过程, 以及预测学生在下一时刻作答的结果. 由于已有的神经网络知识追踪模型在效果和性能上还有待提升, 提出了基于堆叠门控循环单元(Gated Recurrent Unit, GRU)的深度残差(Stacked-Gated Recurrent Unit-Residual, S-GRU-R)网络. 针对长短期记忆网络(Long Short-term Memory, LSTM)参数过多导致过拟合问题, 用GRU代替LSTM学习做题序列中的信息, 采用堆叠GRU扩大序列学习容量, 并用残差连接降低模型训练的难度. S-GRU-R在数据集Statics2011上进行了实验, 并用AUC (Area Under the Curve)和F1-score作为评估指标. 结果表明S-GRU-R在这2个评估指标上都超过了其他类似的循环神经网络模型.

关键词: 深度学习, 知识追踪, 循环神经网络, 门控循环单元, 残差网络

Abstract:

The concept of knowledge tracking involves tracking changes in a student’s knowledge level based on historical question records and other auxiliary information, and predicting the result of a student’s subsequent answer to a question. Since the performance of existing neural network knowledge tracking models needs to be improved, this paper proposes a deep residual network based on a stacked gated recurrent unit (GRU) network named the stacked-gated recurrent unit-residual (S-GRU-R) network. The proposed solution aims to address over-fitting caused by too many parameters in a long short-term memory (LSTM) network; hence, the solution uses a GRU instead of LSTM to learn information on the sequence of questions. The use of stacked GRU can expand sequence learning capacity, and the use of residual connections can reduce the difficulty of model training. Experiments on the Statics2011 data set were completed using S-GRU-R, and AUC (area under the curve) and F1-score were used as evaluation functions. The results showed that S-GRU-R surpassed other similar recurrent neural network models in these two indicators.

Key words: deep learning, knowledge tracking, recurrent neural network, gated recurrent unit, residual network

中图分类号: