Research On Test Item Selection Methods Of Computerized Adaptive Testing

Along with the decades of rapid development of computer and Internet techniques, and for the teaching evaluation there also comes up new realization form, From the traditional paper-and-pencil test to computerized adaptive testing. The idea of CAT is to simulate the intelligent evaluation of human, and by the choice of computer, every student is given suitable test item according to different ability level, and to do more accurate evaluation to the ability level of the tested through statistics of the response of the tested to the item. Thus, CAT could truly achieve to run test according to ability, and provides teachers strong support for teaching according to aptitude.As the central part of CAT, item selection method has a big effect on test efficiency, accuracy, safety, fairness and so on, which makes item selection method very important in CAT.This paper aim to solve the problem of exposure rate unevenness and low efficiency in the item selection method of CAT. the main research work includes the following points:1. This paper through the study of a large number of literature review, comparative analyzed and summarized of several main item selection method,such as Maximum information method,Stratification method,SH method,item selection method with exosure-control factor,points out the problems in the application of various methods.2. This paper puts forward the power function of" ascending a" processing in the whole item bank, and solves the problem that the A-stratification method do not considerate the "ascending a" problem of the layer and layer.3. To control the balance of content, and to solve the problem that the common topic selection strategy does not take into account the impact of content balance.4. In this paper, the particle swarm optimization algorithm is used to solve the objective function, which improves the speed of the selected topic.Last,on the basis of design of experiment, it is found that the proposed new selection method in variable length measurement and long test can better control item exposure rate and speed of the topic has improved greatly, so this method more applicable to the question database safety requirements higher and user experience higher usage scenarios.