| Summary: | 碩士 === 世新大學 === 資訊管理學研究所(含碩專班) === 97 === One of the most important goals of tests designing is to pick items with most discrimination. In the past, most work assumed no dependent relations among test items so that the test papers are made by picking items with highest individual discriminations. But in reality, test items may relate to other items, the overall discrimination of a test paper can not be simply added-up. Hence, this study proposes a two-step method to design test papers by picking discriminative items combinations from the item bank. We first analyze the archival tests to discover substitute items as well as recognize discriminative test itemsets by using data mining technology. Then, the test items are recommended to complete the discriminative test paper.
Finally, a real life case is used to testify the proposed method. These test data are provided by the Chinese Enterprise Planning Association (CERP) in Taiwan. The experimental results show the two-step method can complete the test design task efficiently. In addition, the newly composed test paper presents high discriminative since it is very close to the maximum discrimination under the assumption of items independence which are ideally generated by the Item Response Theory.
|
|---|
