Enhancing Dengue Outbreak Prediction in Bangladesh: A Weighted Average Ensemble Machine Learning Approach

Abstract

Dengue fever is one of the most urgent public health threats in Bangladesh, particularly in cities like Chattogram, where rapid urban growth, poor waste management, and changing climate conditions create a fertile ground for outbreaks. However, most existing dengue prediction models lack contextual adaptation, often relying on single classifiers that fail to capture localized socio-environmental factors, limiting their predictive reliability in the Bangladeshi context. To strengthen early prediction and response, this study introduces a Weighted Average Ensemble Learning (WAEL) model that combines the strengths of Support Vector Machine (SVM), Random Forest (RF), and AdaBoost classifiers. Using a dataset of 199 records and nine attributes collected from a Figshare survey on dengue awareness and prevention, extensive preprocessing steps such as feature selection, cleaning, and imputation were applied to ensure high data quality. Six baseline machine learning models, including Decision Tree, Naïve Bayes, k-Nearest Neighbor, SVM, RF, and AdaBoost, were evaluated, with RF and AdaBoost emerging as the strongest individual performers, each achieving 92.5% accuracy and F1-scores above 0.95. The proposed WAEL approach surpassed all individual models, achieving 93% accuracy, 94% precision, 98.5% recall, and an F1-score of 0.955. These findings demonstrate the advantage of ensemble methods in producing more reliable and context-aware predictions by harnessing the complementary strengths of multiple classifiers. Beyond technical performance, the study offers valuable insights for policymakers and health authorities to identify high-risk areas, vulnerable populations, and key factors driving dengue transmission, ultimately providing a data-driven framework for targeted prevention and early intervention in Bangladesh.