基于理想作答反应构建的非参CD-CAT选题策略

A Non-Parametric CD-CAT Selection Methods Based on Ideal Response

研究提出了两种简单有效且适用于混合计分的非参数选题策略(DWIR和HDWIR),通过模拟研究将DWIR和HDWIR与参数的选题策略比较,结果发现:(1)DWIR和HDWIR算法简单,容易理解,前提条件少,应用条件易满足;(2)DWIR和HDWIR适用性广,不仅可用于01计分和多级计分的题目选题,亦适用于基于参数判别方法和非参数判别方法构建的CD-CAT;(3)DWIR和HDWIR的属性分类准确性高于带曝光控制的选题策略,且选题速度较快;(4)DWIR和HDWIR的曝光控制效果好,且为自带曝光控制的选题策略,无需在选题策略中外加曝光控制,简化选题策略。

Cognitive Diagnostic Adaptive Testing(CD-CAT) can conduct adaptive diagnosis and feedback on students’ knowledge status.In today’s personalized education,CD-CAT has a well development prospect and will play an important role in the future educational practice,so it has increasingly become a research hotspot at home and abroad.Item selection methods has always been one of the most concerned techniques in CD-CAT.Therefore,many researchers have put forward many item selection methods,such as KL,PWKL,SHE,MI and their variants.Although the above parameter selection methods has a good performance in the research,it also has the following shortcomings:First,the selection methods of parameters is generally will make the exposure rate of some items too high and threaten the security of the items bank.Second,the selected item methods are complex,with large computational burden and difficult to understand.In recent years,researchers have developed some simple and effective non-parametric cognitive diagnosis methods.Therefore,this study based on the thought of distance discriminant method,try to developed two kinds of suitable for hybrid scoring non-parametric CD-CAT DWIR and HDWIR item selection strategy.The effects of DWIR and HDWIR were explored through two simulation studies.Study 1:Comparison was made between DWIR and HDWIR and selected item methods without exposure control(PS-KL,PS-PWKL and PS-HKL).Four factors were manipulated:number of attributes(5 and 7),items pool quality(high and low items),examinees(150,300 and 500),item selection methods(PS-KL,PS-PWKL,PS-HKL,DWIR and HDWIR methods).There were 2 × 2 × 3 × 5 = 60 conditions in current study.Among these,item selection methods was within-group variable,and the rest were between-group variables.In addition,the covariates included number of items in the item bank(500 items).The evaluation criteria were pattern correct classification rate(PCCR),attribute correct classification rate(ACCR),exposure rate,test overlap ration(TOR) and item selection of time(Titem).Study 2:Comparison between DWIR,HDWIR and selected item methods with exposure control.In study 2,the number of attributes was determined to 7,the item selection methods was RP-PWKL,DWIR and HDWIR,and other conditions were the same as in study 1.The results show that:(1)DWIR and HDWIR algorithms are simple and easy to understand,with few preconditions and easy to meet application conditions;(2)DWIR and HDWIR have wide applicability.They can be used not only for the item selection of 0-1 scoring and multilevel scoring,but also for the parameter CD-CAT and non-parameter CD-CAT.(3) The attribute classification accuracy of DWIR and HDWIR is higher than that of the parameter selection strategy with exposure control,and the selection speed is faster.(4)DWIR and HDWIR have good exposure control effect,and the selection methods is based on exposure control,so there is no need to add exposure control in the selection methods to simplify the selection methods.In conclusion,the proposed DWIR and HDWIR by this paper had overcame the shortcomings stemmed parameter selection methods from CD-CAT,thus they might expect a good prospect and application.And it provided a kind of new methods and techniques in cognitive diagnosis,which might extended the applicable area.