| Cognitive diagnosis computerized adaptive testing(CD-CAT) inherited the advantages of both cognitive diagnostic(CD) tests and computerized adaptive test(CAT). Selection strategies is a part of CD-CAT, a good selection strategies can estimate the object’s state of knowledge quickly and accurately. Therefore, the study of CD-CAT in selection strategies has attracted both domestic and foreign scholars.So far, five Selection strategies has been studied in CD-CAT, they are random selection strategies, KL selection strategies, PWKL selection strategies, HKL selection strategies and SHE selection strategies. Based on some strategies proposed recently, we elected MPWKL selection strategies, and compare it with other methods based on DINA model. Study one make a judgment by the Pattern Correct Classification Rate(PCCR), Marginal Match Rate(MMR), and Test Overlap Rate(TOR). Based on the first study, the second study studied respectively, each calling the project selection strategies of s, g characteristics, measuring the number of different attributes to call the number of projects and master the number of attributes of different types of subjects sentenced precision were studied. The third study is applied research study, using the CD-CAT system based on "Level-2 English Proficiency Test"(L2EPT), we conducted a comprehensive comparative study of each selection strategies. The results showed that: 1) MPWKL, MI and SHE are the best, and the results of them are similar to each other. 2) KL, PWKL, HKL, MPWKL, MI and SHE method prefers to choose the slipping parameter and guessing parameters values smaller projects. 3) KL, PWKL, HKL, MPWKL, MI and SHE method prefer to choose fewer attributes items to subjects. 4) In general, PWKL, HKL, MPWKL, MI and SHE methods increase judgment precision by the number of attributes mastered by subjects. 5) In real exam condition, MI and SHE method words the best, MPWKL is slightly lower than the previous two methods in exposure control and Criteria of rate. |
PDF Full Text Request