An Intelligent On-line Adaptive Testing System --A Study of Fuzzy Scoring

• Main Authors: KaiLong Hsieh, 謝凱隆
• Other Authors: K. T. Sun
• Format: Others
• Language: zh-TW
• Published: 1999
• Online Access: http://ndltd.ncl.edu.tw/handle/90572419187986951467

碩士 === 臺南師範學院 === 資訊教育研究所 === 87 === The purposes of this study are to construct a fuzzy scoring system and evaluate this system via comparing different assessment methods on the simulation data. In the first place, studied the literatures of scoring and fuzzy inference. Basing on these theory and literatures, built and evaluated the system. In order to build up the rule base of fuzzy inference, collected elementary school teachers’ opinions for scoring via questionnaire. After fuzzy scoring system was built, input the original intellectual subjects’ scores made up by computer simulation to system in order to get the final scores. Then more than 200 elementary school teachers estimate these final scores via questionnaire. Finally, some further suggestions for further researches of multi-evaluation, intelligent on line adaptive testing system are presented. Some results are discussed as follows: 1. The difference between traditional scoring and fuzzy scoring: Uncertainty can not be processed well by the traditional scoring methods. So the results of traditional scoring methods can not be trusted. Fuzzy scoring system uses the fuzzy inference technique that has an excellence performance on processing uncertainty and overcomes the shortcomings of traditional scoring methods. In fuzzy scoring system, teachers can change the assessment rules such that the system is fairer than the traditional scoring methods. 2. The assessment quality of using fuzzy inference on integrated scoring: According to the data from questionnaire, the assessment quality of fuzzy scoring system is better than the methods of weighted average, and it''s the same as the results of average, T scoring and weighted T scoring. System allow the user to change the assessment rules so that it’s more flexible than methods of T scoring, average and weighted average etc. 3. The constructing and limiting of fuzzy scoring system: The original scores are divided into 5 groups during fuzzification process so that the amount of rules is reduced for sampling process. It made the system be constructed easily. However, we can obtain more precise results by extending the range of rules, and this can perform for future works.