Machine learning based tuberculosis (ML-TB) health predictor model: early TB health disease prediction with ML models for prevention in developing countries

• Published in: PeerJ. Computer science Vol. 10; p. e2397
• Main Authors: Karmani, Priyanka, Chandio, Aftab Ahmed, Korejo, Imtiaz Ali, Samuel, Oluwarotimi Williams, Aborokbah, Majed
• Format: Journal Article
• Language: English
• Published: United States PeerJ. Ltd 16.10.2024; PeerJ Inc
• Subjects: Algorithms; Developing countries; Diseases; Good health and well-being; Machine learning; Machine learning (ML); Medical research; Medicine, Experimental; Neural networks; Optimal ML model for TB diagnosis; Pakistan; Rankings; Surveys; Tuberculosis; Tuberculosis (TB) diagnosis; Pakistan; Machine learning (ML); Optimal ML model for TB diagnosis; Good health and well-being; Tuberculosis (TB) diagnosis
• ISSN: 2376-5992; 2376-5992
• DOI: 10.7717/peerj-cs.2397

Tuberculosis (TB) remains one of the top infectious killers in the world and a prominent fatal disease in developing countries. This study proposes a prototypical solution to early prevention of TB based on its primary symptoms, signs, and risk factors, implemented by means of machine learning (ML) predictive algorithms. Further novelty of the study lies in the uniqueness of patient dataset collected from three top-ranked hospitals of Sindh, Pakistan, a self-administered survey patient-records that comprises a set of questions asked by the doctors treating TB patients in real-time. A total of 1,200 survey patient-records were evenly distributed among all three hospitals, viz. ICT Kotri, LUMHS Jamshoro, and Civil Hospital Hyderabad. To develop the required prototypes, the research made use of five distinct benchmark ML algorithms: decision tree (DT), Gaussian naive Bayes (GNB), logistic regression classifier (LRC), adaptive boosting (AdaBoost), and neural network (NN), whose performance was evaluated by considering various performance metrics, ., accuracy, precision, recall, F1 score, and confusion matrix. The experimental results, graphically visualized and systematically discoursed, demonstrate that early detection of TB classifiers, including DT, GNB, LRC, AdaBoost, and NN, attained accuracy rates of 92.11%, 89.04%, 90.35%, 93.42%, and 92.98%, respectively. These results indicate effective diagnosis of TB disease by each implemented ML algorithm.