认知诊断计算机自适应测验中选题策略的新进展-《南京师大学报》（社会科学版）

文章信息/Info

Title:: Progress of Item Selection Methods in Cognitive Diagnostic Computerized Adaptive Testing

Keywords:: cognitive diagnostic computerized adaptive testing; item selection; item exposure; balancing attribute coverage

摘要:: 本文主要从提高测量精确性、控制题目曝光率、平衡测验内容三个方面总结了认知诊断计算机自适应测验的选题策略。文章首先简单回顾以往提出的选题策略,分析这些选题策略的不足,再分别介绍了三种改进方法。它们分别是毛秀珍提出的改善PWKL方法精确性的平均PWKL方法,陈平提出的控制曝光率方法,崇伟峰提出的通过后验概率到0.5的距离来平衡属性测量程度的方法。将三种方法与以往的选题策略比较,结果均显示三种新的选题策略效果更好。最后,作者探讨了认知诊断计算机自适应测验的选题策略中这三方面的平衡问题,以及选题策略以后的发展方向

Abstract:: In cognitive diagnostic computerized adaptive testing(CD-CAT), researchers have to take three aspects into consideration in item selection procedure: how to enhance accuracy of item selection methods; how to control the item exposure; and how to balance the attribute coverage. We reviewed some existing important item selection methods, and analyzed their drawbacks in this article. We also introduced three improved methods: Mao Xiuzhen’s two averaged PWKL, Chen Ping’s exposure control method, and Chong Weifeng’s attribute balancing method which uses the distance between a posterior probability of attribute mastery and 0.5. Compared with the old methods, these three new methods can do better in item selection. In the end, we discussed the relationship between the above-mentioned three aspects and the research direction of item selection methods in CD-CAT.