Predictive models for malaria & TB using ML: health decision support in Africa

This research explored application of advanced machine learning techniques in the development of decision support models for enhanced health informatics in Africa. With a focus on two leading diseases in the continent: Malaria and Tuberculosis, the study exploited eXtreme Gradient Boosting (XGBoost) algorithm to predict malaria incidence in six endemic countries as well as Frequent Pattern Growth (FP-Growth) algorithm coupled with logistic regression to classify Drug-Resistant Tuberculosis (DR-TB) cases. The malaria model accounts for climate variability that has significant effect on malaria prevalence by using climatic data which is integrated into it so as to enhance prediction accuracy and allowing early detection and intervention efforts. The TB model, addressing the challenges of invasive and time-consuming diagnostic methods, identifies hidden patterns in DR-TB symptoms to aid rapid and accurate classification. Performance evaluation using metrics such as (Area Under Curve) AUC of (Receiver Operating Characteristics) ROC, classification accuracy, precision, recall and F1-score demonstrated the superior efficacy of the develoed models compared to existing alternatives. This study aimed at eliminating these life-threatening health concerns in Africa by making well-considered clinical choices that provide knowledge-based decision support system for the care providers, policymakers and health organizations.