The Content Balancing Method for Item Selection in CAT

摘要

Compared with traditional testing, Computerized Adaptive Testing owes incomparable advantages. Such as flexibility, reduce the test length and measurement accuracy. There are some components in CAT, the most one is the item selection algorithm. To perform adaptive test, the most frequently adopted method is based on the maximum information (MI) of items to select the examination questions, with the view to draw the most accurate estimation for tester's capacity. There exists, however, flaws of unbalanced item-exposure as well as unequalled usage of item pool in this method. In this paper, we propose a new item selection algorithm CBIS to solve those problems, and then compare our method with MI method by an experiments. The experiment results are promising.