Computer Engineering and Applications ›› 2021, Vol. 57 ›› Issue (21): 132-139.DOI: 10.3778/j.issn.1002-8331.2010-0287

Previous Articles     Next Articles

Knowledge Graph-Assisted Multi-task Feature-Based Course Recommendation Algorithm

WU Hao, XU Xingjian, MENG Fanjun   

  1. College of Computer Science and Technology, Inner Mongolia Normal University, Hohhot 011500, China
  • Online:2021-11-01 Published:2021-11-04



  1. 内蒙古师范大学 计算机科学技术学院,呼和浩特 011500


When collecting online learning information, there is a general lack of data, which makes the recommendation of course resources may be unsatisfactory due to sparse data. In order to solve the above problems, this paper proposes a Multi-Layer Knowledge graph Recommendation(MLKR) algorithm based on an end-to-end deep learning framework that incorporates knowledge graphs. Based on multi-task feature learning, the knowledge graph is embedded in the tasks. Higher-order connections between potential features and entities are established by cross-compression units between tasks to build recommendation models. Accurate recommendation of course resources based on learners’goals, interests and knowledge levels is realized. The experimental results show that the training time and prediction accuracy of MLKR recommendation algorithm are better than the collaborative filtering algorithm and logistic regression model based on users or items, and it has certain application value in the field of course resource recommendation.

Key words: knowledge graph, collaborative filtering, recommendation algorithms, educational big data, Multi-Layer Knowledge graph Recommendation(MLKR) algorithm


在采集在线学习信息时,普遍存在数据缺失的情况,使得课程资源推荐时可能因数据稀疏导致推荐的效果不理想。为了解决上述问题,基于端对端的深度学习框架,提出了融合知识图谱的多任务特征推荐算法(Multi-Layer Knowledge graph Recommendation,MLKR)。基于多任务特征学习,在任务中嵌入知识图谱;在任务之间通过交叉压缩单元建立潜在特征和实体之间的高阶联系,从而建立推荐模型。实现了基于学习者目标、兴趣、知识水平的课程资源精准推荐。实验结果表明,MLKR推荐算法训练时长和预测准确率均优于基于用户或物品的协同过滤算法和逻辑回归模型,在课程资源推荐领域具有一定的应用价值。

关键词: 知识图谱, 协同过滤, 推荐算法, 教育大数据, 多任务课程推荐(MLKR)算法