一种基于自注意力深度知识追踪和协同过滤的C++教学辅助方法

A C++Teaching Aid Method Based on Self-attention Deep Knowledge Tracking and Collaborative Filtering

为了帮助C++教学实践,根据蚌埠学院网络习题平台积累的学生答题数据集,提出一种基于自注意力深度知识追踪和协同过滤的C++教学辅助方法SAINT-Lite-CF,对C++教学预警和个性化习题推荐进行研究。使用裁剪后的SAINT-Lite方法进行知识追踪,获取学生知识水平矩阵进行教学预警,帮助教师及时调整教学方向。利用预测的学生答题正确概率,基于协同过滤进行个性化习题推荐,帮助学生巩固掌握不牢固知识。实验证明SAINT-Lite知识追踪方法对于小数据集有较好性能,后续的习题推荐结果也都符合期望。同时,教学实践表明SAINT-Lite-CFC++教学辅助方法有很好的可信度和实用性。

In order to help C++teaching practice,a C++teaching aid method SAINT-Lite-CF based on self-attentive deep knowledge tracking and collaborative filtering is proposed for C++teaching alert and personalized exercise recommendation based on the student answer data set accumulated in the online exercise platform of Bengbu University.The tailored Saint-Lite method is used to track knowledge,obtain students’knowledge level matrix for teaching warning,and help teachers adjust teaching direction in time.Based on the predicted probability of students’correct answers,personalized exercises are recommended based on collab⁃orative filtering to help students consolidate their grasp of shaky knowledge.Experimental results show that the Saint-Lite knowl⁃edge tracking method has good performance for small data set,and the recommended results of subsequent exercises also meet ex⁃pectations.At the same time,the SAINT-Lite-CF C++teaching aid method proposed in this paper has good credibility and practi⁃cability.