Fast And Accurate Item Selection Strategies Of Computerized Adaptive Test For Cognitive Diagnosis

A new item selection strategy in computerized adaptive testing with cognitive diagnosis(CD_CAT) is proposed in this paper to solve the problem that the size of item bank restricts the speed of item selection during the course of testing . The new method which calculates the expected pattern match rate (PMR) directly . The experimental results show that this new method not only improves the measurement accuracy, but also increases the speed of item selection a lot. The new item selection strategy is achieved as follow: Firstly the item pool is partitioned to some sub-pools according to the attribute pattern contained in the items , and the sub-pool parameters calculated as the mean of the items in the sub-pool before testing. Secondly, the "best" item set which based on the expected PMR is searched ; lastly , a item randomly or a item which has the highest rate of Shannon's Entropy from"best"item set for test in next step is selected .It increases the speed of item selection and keeps or improves the PMR by using the new strategy.