基于双流结构和多知识点映射结构改进的深度知识追踪模型

Deep Knowledge Tracing Model Improved with Dual-stream Structure and Multi-knowledge Points Mapping Structure

目前,知识追踪已成为自适应个性化辅助学习的研究热点,而基于循环神经网络的深度知识追踪(Deep Knowledge Tracing,DKT)模型在知识追踪领域已取得了较好的效果。但是,DKT模型在融合领域特征时仍存在特征消减和知识点关联关系遗忘等问题,其精准性有待提高。为此,文章在梳理DKT模型融合领域特征相关研究现状的基础上,提出了一种基于双流结构和多知识点映射结构改进的深度知识追踪模型,并通过实验验证了此模型的精准性相较于原始DKT模型及其相关的改进模型有明显提升,并指出其在智慧学习环境下学生认知结构刻画和学习服务精准推荐方面具有的广阔应用前景。通过研究,文章旨在提升深度知识追踪的精准性并进一步助力自适应个性化学习的实现。

At present,knowledge tracing has become the research hotspot of adaptive personalized assisted learning,and the deep knowledge tracing(DKT)model based on the recurrent neural network has achieved good results in the field of knowledge tracing.However,the DKT model still has problems such as feature elimination and forgetting knowledge point association relationships when fusing domain features,and its accuracy needs to be improved.Therefore,on the basis of combing the research status related to fusing domain features of the DKT model,this paper proposed a deep knowledge tracing model with dual-stream structure and multi-knowledge points mapping structure(DKTDM),and through experiments verified that its accuracy was significantly improved compared to the original DKT model and its related improved models.In addition,it was pointed out that its application in characterizing students’cognitive structure and accurate recommendation of learning services in smart learning environments would be promising.Through the study,this paper aimed to improve the accuracy of deep knowledge tracing,and contribute more to the realization of adaptive learning.