Research On Item Selection Strategy To Improve The Security Of Item Bank Under Three Computerized Adaptive Test Models

Computerized adaptive test is based on the item response theory,and according to a computer by examinee ability level automatically choose item,Eventually to the examinee ability estimate of a new type of test. In the process of test, selection strategy is the key part, safe and efficient selection strategy is always the key point of the research, this relates to the security of item pool and the accuracy of the test.In the existing selection strategy based on three parameters logistic model,Cheng et al put forward the dynaMFI a-hierarchical selection strategy that can improve the security of the test, structuring by the information of the test accumulating. In this paper using Cheng et al the basis of the introduction of the exposure factor, Try to curing the power function of item discrimination, reduce the influence of discrimination on the information function,and combined with the mean inequality, under the 0-1 score puts forward three new selection strategy. At the same time, this thought is applied to Multi-level Scoring Models of the Grade Reaction Model and Generalized Partial Credit Model, Simulation results show that the new selection strategies maintain the accuracy of the original test, greatly reduce the exposure rate of the item, improve the utilization ratio of item pool and the security of test.