DTKT:An Improved Deep Temporal Convolutional Network for Knowledge Tracing
Ruixin Ma, Lin Zhang, Jingjing Li, Biao Mei, Yunlong Ma, Hongyan Zhang · 2021
In adaptive education, knowledge tracking predicts the future performance of students based on their historical learning trajectory, the modeling of students’ knowledge state is achieved. Through knowledge tracking, personalized learning can be carried out well to promote the future level of education. At present, lots of Knowledge Tracking models can promote the accuracy of prediction by simulating the complex learning process of human beings. However, when faced with intricate data containing a large number of problem types, it cannot perform satisfactorily, and this is exactly the situation that will always be encountered in our real life. Therefore, we introduced Time Convolution Network into the field of knowledge tracking for the first time, which is expert in processing time series prediction tasks. DTKT completes the parallel processing of data through the characteristics of convolutional network, which greatly enhances the computing power. concurrently, and our model uses a novel data processing method, which takes more into account the authenticity and universality of the input data, which can handle more intricate data better than existing models. Eventually, extensive experimental results on various datasets show that our model performance is better than the latest knowledge tracking model, and it also achieves surpassing in speed.