基于GP-DINA模型的学生多级评分的广义认知诊断模型研究

Research on GP-DINA model based generalized cognitive diagnosis model for students′ multi-level scoring

针对目前的多数认知诊断模型不能适用于多级评分的问题,文中基于GP-DINA模型,通过引入潜变量与滑动参数,提出一种面向学生的多级评分广义认知诊断模型。该模型将Q矩阵中的各属性设置为相等权重,并定义滑动矩阵导出相应的模型表达式。通过3 000名被试的35道题共15次循环模拟,发现该模型在线性型、无结构型、发散型、收敛型和独立型5种属性层级结构下均具有较高的模式准确率与边际判准率,能够为类似的认知诊断模型研究提供参考。

In allusion to the problem that most current cognitive diagnosis models cannot be applied to multi-level scoring,a GP-DINA model based generalized cognitive diagnosis model with multi-level scoring for students is proposed by introducing latent variables and sliding parameters. In this model,each attribute in Q matrix is set as equal weight,and the corresponding model expression is derived by defining the sliding matrix. The total simulations of 15 cycles with 35 items from 3000 subjects is carried out,and the results show that the model has high accuracy of pattern and marginal judgment under the five attribute hierarchy strictures of linear form,unstructured form,divergence form,convergent form and independent form,which can provide reference for the similar cognitive diagnosis model studies.