Summary: | 碩士 === 國立臺南大學 === 數位學習科技學系碩士班 === 96 === Currently, the test items selection theory can be divided into two, one is the Classical Test Theory and the the other is Modern Test Theory. The latter is based on Item Response Theory. In the educational measurement, to select items for constructing tests with approximating a target test information function is a very important and difficult task. Compared with the modern approaches of item selection, the traditional ones are simple but the error of the estimated value is large. In this thesis, we propose a new items selection method based on the Hybrid Taguchi-Genetic Algorithm, which may greatly reduce the difference of information between the constructed test and the desired test. The applications of Hybrid Taguchi-Genetic Algorithm to the educational measurement and this technique can be usefully applicable to the items selection strategy.
|