Research on computerized adaptive testing system based on IRT

As a theoretical basis of computerized adaptive test, item response theory (IRT) has been the focus point of research on test theory of education at present and broken through the limitation of the classical test theory. The ability of examinee is related with his response in examination, and all of them are parameterized by IRT. At present, the research on computerized adaptive testing (CAT) based on IRT is limited in lonely study for dichotomous scoring model alone or multi-scoring model. A mixed-scoring CAT system based on the three-layered B/S model structure is developed in this thesis. The research on mixed-scoring model, which both including dichotomous scoring and multi-scoring, is still in development. This research tries to research on mixed-scoring CAT model based on two-parameter logistic model, and give a resolving method. The functional model and structure of the system is provided. The structure of CAT system and the process of test implement are described in details including the module of the system, the structure of the item-bank, the estimate of parameter, the test flow and so on. The test results show that this system has a better performance in calculating the ability of examinee and the test efficiency. The exposure of items is controlled, and the exposure excess of items is reduced. The ratio of superposition in test is lower, and the security of test is better.