【作者】 周婕；

【导师】 丁树良；

【作者基本信息】 江西师范大学 ， 计算机应用技术， 2006， 硕士

【摘要】 计算机化自适应测验(CAT)，是计算机辅助教学及现代测量研究中的一个重要领域，它最大的特点是根据考生能力不同而由计算机从题库中智能选取难度与其能力相匹配的试题施测，但传统的CAT在测试结束后往往只给被试一个得分，而我们测试的目的不仅仅是为了最后的一个得分，而是希望通过测试了解被试在哪些方面掌握了，哪些方面又未掌握，然后有针对性地对被试进行补救教学。显然，传统CAT是做不到这点的，因此我们引入认知诊断一一对被试认知结构进行诊断分类，考察被试在所测知识点是否掌握，这对教、学双方都有指导作用。目前，认知诊断已经成为CAT研究的重要部分之一，但目前的研究大多都基于0—1评分模型，这与我国测试题目(如证明题，计算题)的多级评分现状不相适应。因此，研究基于多级评分模型的认知诊断具有重要的意义。 本文主要研究开发多级评分模型下带有认知诊断功能的CAT测试系统，采用等级反应模型(GRM)，使用先认知诊断后估计能力的做法：考虑被试掌握属性模式和项目所需属性模式之间的对应关系，对被试作答反应得分进行加权处理，根据被试反应对其掌握模式进行分类，确认被试认知状态后继续对其能力进行估计，直到达到一定信息量则结束测验。模拟实验结果表明：该CAT的认知诊断正确率较高，且诊断结果较稳定，对被试的能力估计也较准确，测验安全性好。本文的创新点如下： ● 研究开发基于多级评分模型下的CAT测验，结果表明，本测验能力估计准确稳定，测验安全性好，可投入实际测验。 ● 在CAT中加入认知诊断，结果表明，该方法对各属性诊断准确率均较高，但对被试属性结构的诊断准确率还有待进一步提高。更多还原

【Abstract】 Computerized Adaptive Testing (CAT) is an important part of the computer assisted instruction and modern measurement research , the aim of CAT is to construct an optimal test for each examinee, this is realized by estimating the examinee’s ability after administration of each item and selecting the next item based on the current ability estimate. However, the traditional CAT only gives a score to the examinee after the test. But the information about which aspects the examinees have mastered or have not can’t be reported by the traditional CAT. Recently, the cognitive diagnosis has become one of important parts of research on CAT ,but most of the researches are based on dichotomous scoring model, the extent of which is not applied extensively.So the cognitive diagnosis under polychotomous scoring model is very meaningful.Cognitive diagnosis CAT based on Polychotomous Scoring Model has been researched and developed in this paper. the examinee’s ability has been estimated after carrying out cognitive diagnosis under the Samejima Graded Response Model: the corresponding relationship between examinee’s attribute mastery patterns and item’sattribute mastery patterns was taken into account. And examinee’s response was dealt with weightedly. Then classify the examinees according to their response rector. The examinee’s ability was estimated after confirming their cognitive situation. The CAT can’t finish until the given amount of information was achieved. The results of this design indicate that accuracy of the CATs diagnosis is comparative higher andsteadier, the ability of evaluation was also more accurate and safer.The main creative points of this paper are: CAT based on Polychotomous Scoring Model has been researched and developedin this paper. The experimental results show that the examinee’s ability of evaluation was more accurate and the CAT was safer.更多还原