|
The purpose of this research is to improve the efficiency of item selection strategy in traditional CAT (Computerized Adaptive Testing) through Gray Prediction Method (GPM) . Based on the Gray Theory, student's competence can be predicted through his/her previous responses and feed back a new item with maximum information among item pool to proceed competence evaluating. The calculation of maximum information is carried out with three parameter method based on Item Response Theory. The prediction operation will be continued until the student's ultimate competence is reached. It was found that the convergence speed of competence prediction of GPM is faster than that of traditional CAT method after several experiments. That is, student's competence can be found within fewer tested items through GPM than traditional CAT method with the same error tolerance. It is apparent that the strategy of GPM can be an efficient method for selecting a proper item in CAT.
|