Paper Title
The Development Of Information System For Multiple-Choice Item Analysis And The Development Of Test Bank

The objective of this research is the development of the information system for multiple-choice item analysis and the development of test bank by using the flexible optimization for unsupervised learning (FOUS) technique. There are three components in the FOUS technique, which are the exam selection, the exam management and reporting exam results. The examination was focused on the assessment of information technology subjects in order to measure the quality and reliability of the test suite. The research methodology is a development of online assessment application based on PHP and to use SQL Server for recording all data. In addition, the FOUS technique uses many algorithms for increasing performance such as the heuristic search, decision tree, rule base, and fuzzy logic. The data was collected from studentís examination and was used to analyze for quality evaluation of the examination sets in 3 aspects, which are reliability, difficulty and discrimination. After that, the system shows the result in graphical and tabular forms. The instructor can change the random criteria according to the test difficulties for improving more appropriate new test suite. There are 100 multiple-choice items for testing the FOUS technique. The results can be concluded that 83 multiple-choice items can be classified by difficulty levels and discrimination levels but 17 multiple-choice items cannot be classified from these level. Then this research solves 17 multiple-choice items by fuzzy logic. However, 74 multiple-choice items pass evaluation criteria and are picked up in the test bank. These multiple-choice items are used to test by 450 students in grade 2 from Suankularbwittayalai Samutprakanschool, which is a popular school in Thailand. These students are divided 2 category, which are the children learn well (average score is more than 2.75) and children learn soft (average score is less than 2.75). This test concludes that 97 percent of the children learn well can do simple tests while 62 percent can do difficult tests. In addition, 76 percent of the children learn soft can do simple test while 22 percent can do difficult tests. It can be seen that the FOUS technique can be classified as higher than difficulty levels and discrimination levels that is 17 items. Indexterms: Assessment test online, Heuristic search, Decision tree, Rule base, Fuzzy logic.