F-TCKT:融合遗忘因素的深度时序卷积知识追踪模型

F-TCKT:deep temporal convolutional knowledge tracking model with forgetting factors

智慧教育中,对学生的知识水平进行追踪是很重要的技术之一。传统的深度知识追踪方法的主要关注点集中在循环神经网络(recurrent neural network, RNN)上,但RNN存在梯度消失或者梯度爆炸的问题,并且很多知识追踪方法没有考虑到学习过程中遗忘行为对结果的影响。针对以上问题,为了准确地预测学生的知识水平,提出了一种融合遗忘因素的深度时序卷积知识追踪模型(temporal convolutional knowledge tracking with forgetting, F-TCKT)。该模型引入了三个影响学生遗忘行为的因素,包括学习相同知识点的时间间隔、学习的时间间隔和同一知识点的学习次数。首先利用全连接网络计算得到表示学生遗忘程度的向量并与学生的答题记录进行拼接,然后使用梯度稳定的时间卷积网络(temporal convolutional network, TCN)和注意力机制预测学生下一次答题正误的概率。经实验验证,与传统方法相比,F-TCKT具有更好的预测性能。

Tracking students’knowledge level is one of the most important techniques in intelligent education.Traditional deep knowledge tracking methods mainly focus on RNN,but RNN has the problem of gradient disappearance or gradient explosion.And many knowledge tracking methods do not take into account the impact of forgetting behavior on the results of lear-ning.Aiming at the above problems,in order to accurately predict the knowledge level of students,this paper proposed a deep temporal convolutional knowledge tracking model,which integrated forgetting factors(F-TCKT).This method introduced three factors that affected students’forgetting behavior,including the time interval of learning the same knowledge point,the time interval of learning and the times of learning the same knowledge point.Firstly,the method used the fully connected network to calculate the vector that represented the degree of forgetting of students,and then concatenated the vector with the answer record of students.Finally,the method used the TCN and attention mechanism to predict the probability of students’next answer.Experimental results show that F-TCKT has better prediction performance than traditional methods.