Questions Classification Based on Revised Bloom's Taxonomy Cognitive Level using Naive Bayes and Support Vector Machine

Abstract

Education is an essential aspect in building the social value and norm to produce individuals who can think in high order thinking through learning and teaching activities. As technology keeps growing, an online learning platform has emerged. This platform is called e-Learning. e-Learning allows teachers to save many questions into the e-Learning question bank. However, these questions need to be reviewed so the questions can be matched with the achievement of competence. One educational identification standard that is often to improve the quality of the questions is Bloom's Taxonomy. Bloom's Taxonomy was created in 1956 and revised in 2001. This study compares the performance of the Support Vector Machine and Naïve Bayes algorithms to classify quiz questions based on the cognitive level of Revised Bloom's Taxonomy. In this study, the dataset received two treatments in handling the imbalanced class. One dataset is using SMOTE method, and one another is not using any oversampling methods. The result shows that classification with oversampling datasets had better results than those without oversampling. The Support Vector Machine algorithm with SMOTE has the highest accuracy of 98%, rather than the Naïve Bayes algorithm with SMOTE has an accuracy of 91%.