深度知识追踪模型综述和性能比较

Review and Performance Comparison of Deep Knowledge Tracing Models

知识追踪是一种重要的认知诊断方法,往往被用于在线学习平台、智能辅导系统等信息化教学平台中.知识追踪模型通过分析学生与课程作业的交互数据,即时模拟学生对课程知识点的掌握水平,模拟的结果可以用来预测学生未来的学习表现,并帮助他们规划个性化的学习路径.在过去20多年中,知识追踪模型的构建通常基于统计学和认知科学的相关理论.随着教育大数据的开放和应用,基于深度神经网络的模型(以下简称“深度知识追踪模型”)以其简单的理论基础和优越的预测性能,逐渐取代了传统模型,成为知识追踪领域新的研究热点.根据所使用的神经网络结构,阐述近年来代表性深度知识追踪模型的算法细节,并在5个公开数据集上对这些模型的性能进行全面比较.最后讨论了深度知识追踪的应用案例和若干未来研究方向.

Knowledge tracking is an important cognitive diagnosis method, which is often used in digitalized education platforms such as online learning platforms and intelligent tutoring systems. By analyzing students’ interactions with course assignments, knowledge tracing models can simulate their mastery level of knowledge concepts in courses in real time. The simulation results can be used to predict students’ future learning performance and help them plan personalized learning paths. In the past 20 years, knowledge tracing models have been constructed based on theories of statistics and cognitive science. With the openness and application of educational big data, models based on deep neural networks, referred to as “deep knowledge tracing models”, have gradually replaced traditional models due to their simple theoretical foundations and superior predictive performances, and become a new research hotspot in the field of knowle dge tracing.According to the neural network architectures, the algorithm details of recent representative deep models for knowledge tracing are illustrated, and a comprehensive performance evaluation of the models on five publicly available datasets is conducted. Final ly, some use cases and future research directions of deep knowledge tracing are discussed.