基于学习者能力的注意力机制知识追踪方法

基于学习者能力，针对基于循环神经网络（RNN）和长短期记忆（LSTM）网络的深度知识追踪（DKT）算法对早期知识点关注的不足，提出一种加入注意力机制的DKT算法，并用时隙聚类的方法对不同能力学习者动态分组并赋予不同的注意力权值，以建立更平衡、更客观的知识记忆程度权重分布模型.常用公开数据集上的实验结果表明：该模型优于2种基准模型和2种消融实验模型，说明所提出的模型能更好地表现学习者的知识状态.

Knowledge tracing method of attention mechanism based on learner ability

Based on learner ability, a deep knowledge tracing (DKT) algorithm was proposed with attention mechanism, aiming at the lack of attention paid to early knowledge points by DKT based on recurrent neural network(RNN) and long-short term memory(LSTM) network. A more balanced and objective weight distribution model of knowledge memory degree for learners with different abilities was established by time slot clustering method. The experimental results on common public datasets showed that this model proposed was superior to two benchmark models and two comparison models, which demonstrated that the model proposed in this paper could represent the state of learner knowledge better.