Implementation of Ensemble Machine Learning with Voting Classifier and CLAHE Pre-Processing for Reliable Tuberculosis Detection

Tuberculosis is an infectious disease caused by microorganisms. Tuberculosis is spread through the air and saliva that contain mycobacterium tuberculosis. If not treated immediately, it can disperse to other vital organs, such as the heart and liver, and can even lead to death. In this study, developed a severe tuberculosis detection system using 4200 dataset consist from 3500 Normal and 700 TB from as a combination from three datasets, NLM, Belarus, and RSNA datasets. This study aimed to create lightweight computation with Voting Classifier in Ensemble Learning as the classifier using Imbalance data and Contrast-Limited Adaptive Histogram Equalization (CLAHE). Initial experiments used single machine learning with the best-performing models, Support Vector Machine (SVM), and Random Forest (RF) as classifiers produce an accuracy of 98.6% and 98%, respectively. Both methods were combined using Ensemble Learning without feature extraction. The accuracy, AUC, Recall, Precision, and F1-score using the voting classifier were 99.16%, 99.9%, 98%, 99.3%, and 98.78%, respectively. Apart from high accuracy, the proposed method can also overcome imbalanced data. These results compete with previous research using deep learning.