融合习题难度和遗忘行为的深度知识追踪模型  

Deep knowledge tracking model integrating exercise difficulty and forgetting behavior

作　　者：马芳兰 朱昶胜[2] 朴世超 MA Fanglan;ZHU Changsheng;PO Shichao(Institute of Sensor Technology,Gansu Academy of Sciences,Lanzhou 730000,China;School of Computer and Communication,Lanzhou University of Technology,Lanzhou 730050,China)

机构地区：[1]甘肃省科学院传感技术研究所,甘肃兰州730000 [2]兰州理工大学计算机与通信学院,甘肃兰州730050

出　　处：《甘肃科学学报》2025年第1期8-15,共8页Journal of Gansu Sciences

基　　金：甘肃省青年科技基金计划项目(21JR11RA217);甘肃省科学院优秀青年基金项目(2023YQ-03)。

摘　　要：深度知识追踪是实现在线教育个性化的关键技术,但是目前的深度知识追踪模型普遍存在只考虑问题及其结果,忽略了学习者作答结果的其他因素的影响,导致深度知识追踪模型的可解释性差、预测准确率低等问题。因此,提出了一种融合习题难度和遗忘行为的深度知识追踪模型(FDKT-ED),该模型建立在传统DKVMN模型的基础上,综合考虑作答结果和习题难度的同时,优化模拟学习者学习过程,将遗忘这一关键行为考虑在建模过程中。通过对比实验结果发现,该模型一方面提升了学习过程中的可解释性,能够将知识状态的变化展现出来,另一方面将预测结果准确率提升了2%~4%,预测效果提升明显。Deep knowledge tracking is the key technology to realize the personalization of online education.However,the current deep knowledge tracking model generally only considers the problems and results,while ignoring other factors affecting learners'answer results,which will lead to the poor interpretability and low prediction accuracy of deep knowledge tracking model.Therefore,this paper proposes a deep knowledge tracking model integrating exercise difficulty and forgetting behavior(FDKT-ED).The model is based on the traditional DKVMN model.While comprehensively considering the answer results and exercise difficulty,it optimizes the learning process of simulated learners and considers the key behavior of forgetting in the modeling process.Through the comparison of experimental results,it is found that on the one hand,the model improves the interpretability in the learning process and can show the changes of knowledge state.On the other hand,the accuracy of prediction results is improved by 2%~4%,and the prediction effect is significantly improved.

关 键 词：知识追踪 深度学习 习题难度 遗忘行为 学习过程 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程] G434[自动化与计算机技术—控制科学与工程]