Predictive Adaptive Test with Selective Weighted Bayesian Through Questions and Answers Patterns to Measure Student Competency Levels

Abstract

Computer Assisted Testing (CAT) system in Indonesia has been commonly used but only to displaying random exam questions and unable to detect the maximum performance of the test participants. This research proposes a simple way with a good level of accuracy in identifying the maximum ability of test participants. By applying the Bayesian probabilistic in the selection of random questions with a weight of difficulties, the system can obtain optimal results from participants compared to sequential questions. The accuracy of the system measured on the choice of questions at the maximum level of the examinee alleged ability by the system, compared to the correct answer from participants gives an average accuracy of 75% compared to 33% sequentially. This technique allows tests to be carried out in a shorter time without repetition, which can affect the fatigue of the test participants in answering questions.