具有认知诊断功能的区块化自适应测验的提出及验证

CD-CAT以Q矩阵理论为基础,随着属性个数的增加,理想掌握模式的数量以指数的形式增加,这大大提高了计算难度,计算速度缓慢,判准率大幅降低。为了解决该问题,本文提出了具有认知诊断功能的区块化自适应测验(CD-BAT)。CD-BAT将单个的CD-CAT作为区块进行处理,把前一区块的信息作为先验信息用于后一区块的估计过程,并不断更新和修正前一区块的估计结果。模拟研究表明,CD-BAT的判准率略优于CD-CAT,题库质量下降时,这种优势逐渐加大。随着CD-BAT的区块的增加,其属性判准率呈现上升趋势。

Propose and Verify of Cognitive Diagnostic Blocked Adaptive Testing

Abstract Cognitive Diagnostic Assessing (CDA) aims to determine whether or not examinees have mastered attributes or skills underlying according to their responses to test items, it provides a more detailed evaluation of the strengths and weaknesses of students. It aims to remedial teaching, which can make students understand own knowledge state in-depth, realize self-learning. It also can achieve “teaching students in accordance with their aptitude” which is a thousands of years dream when teachers conduct targeted teaching according to the specific ability of the students. CD-CAT is one of the most successful application. It has the characteristic of cognitive diagnosis and adaption at the same time. It can evaluate the knowledge state of candidates in a short time with fewer questions.Due to classification of the ideal master patterns, the existing CD-CAT could not handle too many attributes. The number of ideal master patterns increases exponentially with the raise of the attribute number. For example, when the attribute number is 4, the category of ideal master pattern is 24=16, when the attribute number increases to 8, the category of ideal master pattern is 28=256. The attribute number is controlled in 4-8 within a unit in teaching practice. When the unit number increases, the attribute number is always more than 10, the existing CD-CAT could not handle under the granularity attribute level. The module of CD-CAT could be considered as a block, CD-MST is a few blocks. CD-CAT is adaptive at item level, it can be seen as one block. This study combined CD-MST and CD-CAT, presented a new adaptive testing—Cognitive Diagnostic Blocked Adaptive Testing (CD-BAT), which adopted dynamic and unique ideal master patterns to each examinee to control the issue of traditional CD-CAT attribute number limited. The simulation study showed that when the attribute number was 8, CD-BAT had slightly advantages than traditional CD-CAT. The reason was that the CD-BAT revised the attribute after the first block. When the attribute number was 12, the PCCR of CD-BAT was deceased, the ACCR appeared an upward trend. In other words, due to the increasing of the attribute number, the accurate of the whole pattern was decreasing, the accurate of the attribute was increasing.