Exploring key risk factors in dengue forecasting with XAI interpreting geographical, test, and demographic facts

Dengue fever is a vector-borne disease spread by mosquitos that has dangerous public health consequences, particularly in tropical areas. Traditional machine learning models such as SVM, DT, and RF are known as black-box models due to their restricted interpretability, which makes it difficult to understand how predictions are created. This study intends to address this gap by improving model dependability and interpretability using explainable AI techniques. SHAP and LIME are used to evaluate a Kaggle dataset with clinical and geographical variables from Dhaka. These techniques provide both global and local explanations, highlighting important risk factors such as NS1, IgG, undeveloped areas, and dwelling styles. This study shows that those who reside in undeveloped areas as well as those who live in tin-shed dwellings are more vulnerable to dengue and the Random Forest (RF) model achieved 86% accuracy at best here. Those aged 10 to 30 are the most vulnerable especially those with elevated IgG levels, while main clinical indicators such as NS1 and IgG have been recognized as mandatory for early detection. These findings suggest that focused treatments can successfully minimize dengue epidemics.