PREDICTING PARTICIPANT COMPETENCE TEST RESULT USING MACHINE LEARNING APPROACH

Abstract

The 3in1 training for electronics operators aims to deliver competent participants. Therefore, it is necessary to improves the learning curriculum and materials. However, the organizer and teaching staff lack information about the factors that determined the success rate of participant's competence tests. Therefore, this paper aims to build a model for predicting participant competence test result. The data of participant's assessment scores were first collected and prepared as the dataset. The prediction model builds by applying a machine learning approach. These cover the use of ANOVA to ranked the course subjects towards the competency test results (C and NC) and build the prediction model using the Random Forest algorithm. From the results, we found that the competency test results are more affected by the practical subjects rather than theory subjects. From the ANOVA results, the most significant practical subject is the screwing lesson, while for theory subject is 5S-Kaizen. The prediction model obtains an accuracy of 94,6% for 5-subjects and 91,9% for 8-subjects from the original dataset. However, from the precision rate, it was found that the oversampling and hybrid sampling dataset shown better results. This confirms that the resampling technique is working to solve the imbalanced dataset problem.