Design And Implementation Of Computerized Adaptive Test System

Item Response Theory is one of the modern psychological theories and educational measurement, the results of which are widely used in many fields, resulting in the gradual change. The traditional examination, which is taken on fixed-time, on the same papers, based on Classical Test Theory, has changed into Computerized Adaptive Test, based on Item Response Theory. The adaptive test, however, is now mainly applied to some large-scale examinations, such as the GRE, but seldom used in the universities.Computerized Adaptive Test (CAT) is a kind of test which combines the measurement technology and computer technology under the guidance of the Item Response Theory. The main idea of CAT is that testees with different levels can be tested according to their own characteristics. That is to say, when tested, the subjects can choose items, selected from the item bank, which correspond to their equivalent abilities. In other words, proficient students take difficult items, poor ones take easy items, with different standards in order to provide the same accuracy of measurement. This truly reflects the actual level of the test. Compared with the traditional test, CAT takes both the accuracy of the test span and delicacy into consideration. Different subjects are tested in different ways, so test time shorten is shortened; test efficiency is improved, test results are more realistic, Therefore the traditional test problems are overcome completely.In this paper, the author has made a detailed analysis of the computerized adaptive system, including needs analysis, general designing, detailed designing and realization, and focusing two key techniques: selection strategies and the ability estimation. The existing item selection strategies have their own characteristics, so we adopt such a item selection strategy which takes three parameters (discrimination, difficulty and guessing degrees), as well as distribution of the contents into account. There are no fixed ability estimation methods. We have chosen the Bayesian posterior expectation estimation (Bayes Expected a Posterior Estimation, EAPE) based on a variety of methods of analysis to carry out the practice ability estimate calculation and simulation experimental comparison of the methods, so as to design and realize the computerized adaptive system.