区分度与测验进程相匹配的CAT选题策略  

The Item Selection Strategy on Composing the Discrimination with the Test Process in CAT

作　　者：李佳[1] 丁树良[1] 况天昊 LI Jia;DING Shuliang;KUANG Tianhao(School of Computer Information Engineering,Jiangxi Normal University,Nanchang Jiangxi 330022,China;College of Information Engineering,Jiangxi University of Technology,Nanchang Jiangxi 330098,China)

机构地区：[1]江西师范大学计算机信息工程学院,江西南昌330022 [2]江西科技学院信息工程学院,江西南昌330098

出　　处：《江西师范大学学报（自然科学版）》2021年第4期384-389,共6页Journal of Jiangxi Normal University(Natural Science Edition)

基　　金：国家自然科学基金(62067004,62067005,61967009)资助项目。

摘　　要：选题策略是计算机化自适应测验(CAT)的核心.该文提出了一种新的选题策略,是一种相对严格的"升a"方法,它选择区分度参数的百分等级尽可能接近测验进程的项目,而且还可以通过调整控制参数的取值来满足不同测验场景的需求.Monte Carlo实验结果表明:该方法在测验精度、项目曝光率控制和题库利用率等方面均表现良好.The item selection strategy is an important content in computerized adaptive testing(CAT).A new item selection strategy about composing the percentile rank of the discrimination parameter for an item in the bank with the particular examinee′s test process is introduced in this paper.The control parameter can be changed to meet the different test situations.New method has better performance through the Monte Carlo simulation method on improving the test precision,controlling item exposure and the utilization of the item bank.

关 键 词：计算机化自适应测验 选题策略 项目曝光控制 题库利用率 控制参数 

分 类 号：B841[哲学宗教—基础心理学]