中国邮电高校学报(英文) ›› 2023, Vol. 30 ›› Issue (2): 36-48.doi: 10.19682/j.cnki.1005-8885.2022.0020

• Artificial Intelligence • 上一篇    下一篇

KG-based memory recommendation algorithm for learning path

王丹志,向键鑫,崔岩松   

  1. 北京邮电大学
  • 收稿日期:2021-12-20 修回日期:2022-06-19 出版日期:2023-04-30 发布日期:2023-04-27
  • 通讯作者: 向键鑫 E-mail:paoxbin@qq.com

KG-based memory recommendation algorithm for learning path

Zhi DanWANG1,Jian-Xin XIANG1,Yan-Song CUI   

  • Received:2021-12-20 Revised:2022-06-19 Online:2023-04-30 Published:2023-04-27
  • Contact: Jian-Xin XIANG E-mail:paoxbin@qq.com

摘要:

In intelligent education, most student-oriented learning path recommendation algorithms are based on either collaborative filtering methods or a 0-1 scoring cognitive diagnosis model. Unfortunately, they fail to provide a detailed report about the students'mastery of knowledge and skill and explain the recommendation results. In addition, they are unable to offer realistic learning path recommendations based on students'learning progress. Knowledge graph based memory recommendation algorithm (KGM-RA) was proposed to solve these problems. On the one hand, KGM-RA can provide more accurate diagnosis information by continuously fitting the students' knowledge and skill proficiency vector (SKSV) in a multi-level scoring cognitive diagnosis model. On the other hand, it also proposes the forgetting recall degree (FRD) according to the statistical results of the human forgetting phenomenon. It also calculates closeness centrality in the knowledge graph to achieve the recommended recall effect consistent with the human forgetting phenomenon. Experiments show that the KGM-RA can obtain the actual learning path recommendations for students, provides the adjustable ability of FRD, and has better reliability and interpretability.

关键词: learning path recommendation| knowledge graph| cognitive diagnosis| human forgetting phenomenon

Abstract:

In intelligent education, most student-oriented learning path recommendation algorithms are based on either collaborative filtering methods or a 0-1 scoring cognitive diagnosis model. Unfortunately, they fail to provide a detailed report about the students'mastery of knowledge and skill and explain the recommendation results. In addition, they are unable to offer realistic learning path recommendations based on students'learning progress. Knowledge graph based memory recommendation algorithm (KGM-RA) was proposed to solve these problems. On the one hand, KGM-RA can provide more accurate diagnosis information by continuously fitting the students' knowledge and skill proficiency vector (SKSV) in a multi-level scoring cognitive diagnosis model. On the other hand, it also proposes the forgetting recall degree (FRD) according to the statistical results of the human forgetting phenomenon. It also calculates closeness centrality in the knowledge graph to achieve the recommended recall effect consistent with the human forgetting phenomenon. Experiments show that the KGM-RA can obtain the actual learning path recommendations for students, provides the adjustable ability of FRD, and has better reliability and interpretability.

Key words: learning path recommendation| knowledge graph| cognitive diagnosis| human forgetting phenomenon