Computer-based Adaptive Test Development Using Fuzzy Item Response Theory to Estimate Student Ability

Abstract

The field of computing has developed so rapidly. Various theories of computational evolution to support human needs are continually being pursued; one of them is the field of education, especially in terms of teaching, testing, and evaluation of exam results. This study aims to develop computerized adaptive tests (CAT) to measure the student's abilities. Students will be measured for their cognitive abilities in Mathematics and Science subjects. It starts with developing a question bank that has been tested with 720 students to classify items based on its characteristic, i.e., easy, medium, and challenging. This research uses the item response theory approach with the model 2 logic parameters (2PL), namely item difficulty and item difference power. The selection of test items for each participant will depend on the response of the previous answer. Fuzzy algorithm is used in analyzing test items through four stages, namely fuzzification, implications, inference, and defuzzification. Meanwhile, to measure the ability of test-takers, the maximum likelihood estimation method, MLE, is used. Based on the testing of 73 students, it was found that each student received a different test item, both in the number of questions and the level of difficulty of the questions, according to student's abilities. The results of the CAT program's measurement of the test taker's ability estimation were stated to be more effective compared to conventional methods, as indicated by the average test length of 15 items compared to traditional tests, which had a length of 50 items. Therefore, the CAT program with the fuzzy item response theory can be used as support to measure students' abilities.

Analysis of Pre-Olympic Middle School Mathematics Test Instruments Based on Item Response Theory

“…Second, items 7, 11, and 15 have the ability to distinguish students in the medium category. Third, points 3, 8, and 10 are able to distinguish students' abilities in the low category (Wulandari et al, 2020). Furthermore, items 1, 6, and 13 have very low discriminating power.…”

Kalibrasi Instrumen Literasi Matematika Siswa Menggunakan IRT dan Aplikasinya untuk Estimasi Skor

“…The Rasch model only produces difficulty parameters, which can be seen in column 2 of Table 6. Wulandari et al (2020) categorized the level of difficulty into three categories, namely difficult (b ≥ 0.7), moderate (-1 ≤ b < 0.7), and easy (b < -1). The mathematical literacy instrument for algebra content in Table 6 has a level of difficulty at the easy level for items 2 and 15.…”