基于信息量的试题非等权重神经认知诊断  

The Neural Cognitive Diagnostic Model Integrated Item Weight Based on Information Function

作　　者：李梦超 罗芬[1,2] 熊建华[1,2] LI Mengchao;LUO Fen;XIONG Jianhua(School of Digital Industry,Jiangxi Normal University,Shangrao Jiangxi 334000,China;School of Computer and Information Engineering,Jiangxi Normal University,Nanchang Jiangxi 330022,China)

机构地区：[1]江西师范大学数字产业学院,江西上饶334000 [2]江西师范大学计算机信息工程学院,江西南昌330022

出　　处：《江西师范大学学报（自然科学版）》2024年第3期286-293,共8页Journal of Jiangxi Normal University(Natural Science Edition)

基　　金：国家自然科学基金(61967009,62267004)资助项目。

摘　　要：该文提出了基于信息量的试题非等权重神经认知诊断(WNCD)方法,即利用学生能力水平与试题难度之间的相近程度来调整试题在学生认知状态估计时的权重,并通过神经网络拟合试题和学生之间复杂的非线性交互关系.与已有方法在ASSIST和PISA2012数据集上进行对比实验的结果表明:WNCD不仅保持了良好的解释性,而且提升了诊断精度.The neural cognitive diagnostic model integrated item weight based on information function(WNCD)is proposed,which adjusts the weights of test items for inferring student cognitive status based on the proximity estimated between student abilities and item difficulties.The model also utilizes neural networks to capture the complex nonlinear interactions between test items and students.Comparison experiments with existing methods on ASSIST and PISA2012 data sets show that WNCD not only maintains good interpretability,but also outperforms the equal-weight neural cognitive diagnosis models on diagnostic accuracy.

关 键 词：认知诊断 神经网络 试题权重 信息函数 

分 类 号：B841[哲学宗教—基础心理学] O211.67[哲学宗教—心理学]