TCN-KT：个人基础与遗忘融合的时间卷积知识追踪模型

智慧教育的热门领域知识追踪（KT）被视为序列建模任务，其主要关注点和解决方式都集中在循环神经网络（RNN）上。但RNN通常会面临梯度消失或者梯度爆炸等问题，且训练时间和设备要求都过于严苛，针对以上问题，提出融合学习者个人先验基础和遗忘因素的时间卷积知识追踪模型（TCN-KT）。首先利用RNN模型计算得到学生个人先验基础，然后使用梯度稳定、内存占用率更低的时间卷积网络（TCN）预测学生下一题正误的初始概率，最后融合基于学生基础的遗忘因素得到最终结果。实验验证，TCN-KT预测性能最佳并减少了计算时间。

TCN-KT:temporal convolutional knowledge tracking model based on fusion of personal basis and forgetting

KT is a popular area of wisdom education and is a typical sequence modeling task. Its main focus and solutions are focused on RNN. However, RNN''s training time and equipment requirements are too strict, which usually leads to problems such as gradient disappearance or gradient explosion. In response to the above problems, this paper proposed temporal convolutional network knowledge tracing model(TCN-KT) that integrated the learner''s personal prior basis and forgetting factors. Firstly, the method used the RNN model to calculate the student''s personal prior basis. Then, the model used the gradient-stable and lower memory usage TCN to predict the initial probability of the student''s next question. Finally, the model got the final result by integrating the forgetting factors based on the student''s foundation. Experimental results show that TCN-KT has the best performance and reduces calculation time.