An Explainable Machine Learning Framework for Dengue Disease Prediction Using MG-PSO

Dengue fever, transmitted through Aedes mosquito bites, continues to escalate worldwide, placing disproportionate strain on tropical healthcare systems due to its capacity to progress into life-threatening complications. Timely and accurate identification of dengue cases is crucial to prevent disease progression and reduce mortality risks. Machine Learning (ML) has become a potent tool in predictive healthcare due to the quick development of computer intelligence, allowing for accurate disease classification and early decision assistance. In this study, we propose an intelligent and data-driven dengue disease (DD) prediction framework designed to distinguish between Dengue and Non-Dengue cases using supervised ML techniques. A publicly accessible Kaggle dataset is used to assess many categorization models, such as Random Forest (RF), Extreme Gradient Boosting (XGB), Categorical Boosting (CB), Logistic Regression (LR), and Support Vector Machine (SVM). A Memory-Guided Particle Swarm Optimization (MG-PSO) based Feature Selection (FS) method is used to improve predictive performance and find the most pertinent clinical features. The MG-PSO optimized XGB model obtains the maximum accuracy of 99.95% in experimental validation using 10-fold cross-validation, far beating its non-optimized counterpart and other competing models. To guarantee the accuracy and reliability of model forecasts, Shapley Additive explanations (SHAP) and Local Interpretable Model-agnostic Explanations (LIME), two Explainable Artificial Intelligence (XAI) approaches, are used to offer intuitive insights regarding feature contributions. The results show that the suggested architecture provides an efficient, understandable, and highly accurate solution for DD prediction, underscoring its potential use in public health surveillance and clinical decision support systems.