In academic institutions, written exams are the most conventional and traditional method of evaluating students. The need to assess test question quality and students’ cognitive abilities has been highlighted by educational advancements. On the other hand, the professor finds the process of creating questions to be quite difficult. Attempts by lecturers to provide fair and high-quality questions to evaluate students at varying cognitive levels are becoming increasingly difficult. High-quality tests are written, assessed, and measured using Revised Bloom’s Taxonomy (RBT), which is extensively utilized in the educational setting. As a result, numerous scholars have worked on automating the RBT based grouping of test questions. This study suggested using the OPTUNA Hyperparameter Optimization Framework to automatically categorize exam questions based on RBT’s cognitive levels. In order to categorize exam questions according to RBT, the research work suggested an improved Light Gradient Boosting Machine (LGBM) classifier. Based on Term Frequency Inverse Document Frequency (TF-IDF), the Feature Extraction (FE) is performed. In order to categorize the test questions and choose a reasonable weight for the vital words in the question, the FE is to compute and identify the main features from the raw data based on part of speech. With a 95.3% accuracy rate, the proposed hyperOPTUNALGBM classifier outperformed than the other Machine Learning (ML) models.