Malaria risk assessment in Indonesia: a machine and deep learning framework

Abstract

This study focuses on developing comprehensive malaria risk model for Indonesia, integrating susceptibility, vulnerability, and capacity to better understand and manage malaria risks across the country. The primary objective was to identify high-risk areas and prioritize malaria management efforts by combining machine-deep learning techniques and socioeconomic data. Using Gradient Tree Boosting, Classification and Regression Tree, Random Forest algorithms and Deep Learning Multilayer Perceptron, the study analyzed malaria susceptibility, revealing that 38% of Indonesia’s territory was categorized as highly susceptible, with the provinces of Central Kalimantan, West Kalimantan, East Kalimantan, South Sumatra, and Papua identified as the most affected regions. Novel aspects of this study include integrating age and sex ratios to model vulnerability and calculating healthcare access to assess capacity, which showed that 65% of the territory exhibited high vulnerability and 34% had low healthcare capacity, with Kalimantan and Papua consistently ranking highest in risk factors. By combining these factors, the final malaria risk model identified 88 cities with high malaria risk, of which 60 cities with low Gross Regional Domestic Product were prioritized for intervention. This research contributes to malaria control by offering a detailed and data-driven framework to guide policy and resource allocation, enhancing efforts to achieve sustainable health outcomes in malaria-endemic regions.