Computer Engineering and Applications ›› 2021, Vol. 57 ›› Issue (6): 117-123.DOI: 10.3778/j.issn.1002-8331.1912-0276

Previous Articles     Next Articles

Knowledge Tracing Model Based on Mastery Speed

ZONG Xiaoping, TAO Zeze   

  1. College of Electronic Information Engineering, Hebei University, Baoding, Hebei 071002, China
  • Online:2021-03-15 Published:2021-03-12

基于掌握速度的知识追踪模型

宗晓萍,陶泽泽   

  1. 河北大学 电子信息工程学院,河北 保定 071002

Abstract:

Knowledge Tracking(KT) model is a hot topic in the field of educational data mining. KT can find student’s weak knowledge points and recommend the best learning path and exercises to students. The existing KT model is only applicable to the modeling of a single knowledge point and does not consider the problem of students’ mastery speed. Mastery Speed Knowledge Tracing(MSKT) model is proposed. MSKT borrows from the idea of Memory Augmented Neural Network(MANN) and the advantages of Dynamic Key-Value Memory Network model(DKVMN), and uses the current memory contents when calculating delete vectors and add vectors. Experiments show the effectiveness and superiority of MSKT model, and similar exercises can be found automatically.

Key words: educational data mining, knowledge tracing, deep learning, personalized recommendation, Memory Augmented Neural Network

摘要:

知识追踪(Knowledge Tracing,KT)是教育数据挖掘领域的研究热点。KT能够自动发现学生的薄弱知识点,向学生推荐最佳的学习路径和练习题。针对现有的KT模型只适用于对单个知识点建模并没有考虑学生掌握知识点快慢的问题。提出了一种基于掌握速度的知识追踪模型(Mastery Speed Knowledge Tracing,MSKT),MSKT采用了记忆增强神经网络(Memory Augmented Neural Network,MANN)的思想和动态键值记忆网络模型(Dynamic Key-Value Memory Networks,DKVMN)的优点,并且在计算删除向量和增加向量时,使用了当前的记忆内容。通过对比实验验证了MSKT模型的有效性和优越性,并且可以自动地发现相似练习题。

关键词: 教育数据挖掘, 知识追踪, 深度学习, 个性化推荐, 记忆增强神经网络