深度学习支持下的自适应学习路径推荐——融合图嵌入与注意力机制的知识追踪模型

Adaptive Learning Path Recommendation Supported by Deep Learning——Knowledge Tracing Model Integrating Graph Embedding and Attention Mechanism

自适应学习路径推荐是智能技术在教育服务智能化中的核心应用。知识追踪通过分析学生的历史学习记录,预测其未来知识水平,从而为其提供个性化的学习规划与推荐。然而,现有的知识追踪方法在处理学生在线学习行为的复杂性时存在数据稀疏问题,并且忽略了时间因素和学生的遗忘机制,导致模型不能准确捕捉学生的状态变化,影响推荐效果。因此,在知识追踪模型中融合图嵌入和注意力机制,设计出一种新颖的深度学习支持下的自适应学习路径推荐模型(GE-MAKT)。实验结果表明,相较于传统方法,GE-MAKT模型的AUC和ACC两个评价指标得到显著提升,增强了对学生知识掌握水平的判断能力,可为学生提供更加个性化的学习路径推荐。

Adaptive learning path recommendation is a core application of intelligent technology in the intelligentization of educational servic⁃es.Knowledge tracing predicts students'future knowledge levels by analyzing their historical learning records,providing personalized learning plans and recommendations.However,existing knowledge tracing methods face data sparsity issues when dealing with the complexity of stu⁃dents'online learning behaviors and often ignore temporal factors and students'forgetting mechanisms.This leads to models failing to accurate⁃ly capture students'state changes,thus affecting recommendation effectiveness.This paper integrates graph embedding and attention mecha⁃nisms into the knowledge tracing model,designing a novel adaptive learning path recommendation model supported by deep learning(GEMAKT model).Experimental results show that compared to traditional methods,the GE-MAKT model significantly improves the AUC and ACC evaluation metrics,enhancing the ability to assess students'knowledge mastery levels,thereby providing more personalized learning path recommendations for students.