Text Mining Approach Using TF-IDF and Naive Bayes for Classification of Exam Questions Based on Cognitive Level of Bloom's Taxonomy

Bloom's Taxonomy is a unity of three domains, which are divided into lower orders and high orders based on the Bloom Taxonomic Cognitive Domain, the level is used to classify learning objectives and serve as benchmarks for evaluating student achievement. Basically, an evaluation of student achievement can be done by giving questions on exam activities. The questions given are then classified according to the level in the Cognitive Domain. However, because the number of questions is too many and the classification is still manual, it causes the classification results are not accurate and inconsistent. Therefore, the employing of the Naive Bayes Classifier in classifying exam questions based on levels in the Cognitive Domain can be a solution. This study uses real-world dataset collected from mid-terms and final exams questions taken from Department of Information Systems, Telkom University from the academic year 2012/2013 to 2018/2019. In particular, we examined Words, Characters, and N-gram as indexing terms. The results showed that the classification using Naive Bayes and TF-IDF with N-gram as indexing terms achieved precision of 85% and recall of 80%.