Background Malaria remains a significant threat to global health, with substantial economic and public health implications. In 2019, over 247 million cases of malaria were reported, resulting in 619,000 deaths. More than 42 million Brazilians are at risk of developing malaria, with cases concentrated in the Legal Amazon region (WHO, 2020). Malaria continues to be a major public health issue in Brazil.
Methods In this study, we focus on predicting malaria cases in the high-incidence region of Legal Amazon. Our investigation contributes in two main ways: by developing a comprehensive data preprocessing methodology for the SIVEP-MALARIA dataset and by providing the preprocessed dataset to enhance research reproducibility. Subsequently, we employ two machine learning models, SVR and Random Forest, to predict malaria cases in Legal Amazon states. Our analysis spans different states and malaria categories, assessing model performance using RMSE and MAE metrics.
Results The results indicate that the SVR model outperformed for the Vivax and Falciparum Malaria datasets. Random Forest exhibited superior performance in states and malaria categories with limited case notifications, providing accurate predictions even when no notifications were available.
Conclusions Evaluating each state individually and considering each malaria type is an approach for policymakers to tailor public policies according to each state's specific needs. Ongoing actions and innovative approaches are essential in the fight against malaria, where the integration of predictive tools leveraging epidemiological data stands out as a promising strategy to overcome this global challenge. Our findings point towards a potential path to malaria elimination in the Legal Amazon region.