深度学习驱动的知识追踪研究进展综述

随着教育信息化程度的不断加深，以预测学生知识状态为目标的知识追踪正成为个性化教育中一项重要且富有挑战性的任务。知识追踪作为一项教育数据挖掘的时间序列任务，与深度学习模型强大的特征提取和建模能力相结合，在处理顺序任务时具有得天独厚的优势。为此，简要分析传统知识追踪模型的特点及局限性，以深度知识追踪发展历程为主线，总结基于循环神经网络、记忆增强神经网络、图神经网络的知识追踪模型及其改进模型，并对该领域的已有模型按照方法策略归类整理。同时梳理了可供研究者使用的公开数据集和模型评估指标，比较和分析不同建模方法的特点。对基于深度学习的知识追踪的未来发展方向进行探讨和展望，奠定进一步深入基于深度知识追踪研究的基础。

Review of Knowledge Tracing Preprocessing Based on Deep Learning

As education informatization keep deepening, knowledge tracing with the goal of predicting students’ knowledge status is becoming an important and challenging task in individualized education. As a time sequence task of educational data mining, knowledge tracing combines with the powerful feature extraction and modeling capabilities of deep learning models, and it has the unique advantage when dealing with sequential tasks. To this end, this article discusses knowledge tracing from the following four aspects. Firstly, the article briefly analyzes the characteristics and limitations of traditional knowledge tracing models. Then, through taking the development process of in-depth knowledge tracing as the main line, it summarizes knowledge tracing models based on recurrent neural networks, memory augmented neural networks, graph neural networks and their improved models; and categorizes and organizes the existing models in this field according to methods and strategies. Besides, this article sorts out the public data sets and model evaluation indicators which can be used by researchers, compares and analyzes the characteristics of different modeling methods. Finally, it discusses and prospects the future development direction of knowledge tracing based on deep learning, and lays the foundation for further in-depth knowledge tracing research.