A novel approach for dengue outbreak prediction using evolutionary sampling with prediction framework

The increasing prevalence of life-threatening viral diseases like dengue fever necessitates comprehensive research into their causes, recovery, and preventive measures. Dengue outbreak data often suffers from irregularities, underreporting, delays, and missing information, which challenge the development of reliable prediction models.

To overcome these issues, the study proposes an innovative framework that combines Evolutionary Sampling with Prediction (ESP) to handle temporal and stochastic dynamics, along with a Minimax K-nearest neighbour imputer to correct missing data biases. Additionally, a novel Firefly Dynamic Evolution (FDE) approach optimizes model parameters, while a Random Forest classifier captures complex, nonlinear relationships in the data. The model was evaluated using 10-fold cross-validation on two datasets: the Local Epidemics Dengue Fever dataset (San Juan and Iquitos) and the Brazil dengue dataset.

The proposed model achieved a low Mean Absolute Error (MAE) of 22.1 and Root Mean Squared Error (RMSE) of 46.37 on the local dataset, and an MAE of 48.36 and RMSE of 86.76 on the Brazil dataset, demonstrating improved accuracy and robustness.

These findings highlight the model's potential for early warning systems and broader applications in forecasting other infectious diseases.