PERTANIKA JOURNAL OF SCIENCE AND TECHNOLOGY

Dengue, a vector-borne disease, has become a global calamity. A robust early warning mechanism could assist authorities in mitigating this hazard. The objective of the present study is to propose a dengue forecasting model employing a high-throughput dataset for eleven months of dengue progression in the United States of America (USA). The methodology integrates the efficacy of well-renowned statistical methods, i.e., Auto ARIMA, Auto ETS, and machine-intelligent methods, namely Extreme Learning Machine (ELM). The final architecture of the proposed approach is realised by using an evolutionary algorithm, i.e., Genetic Algorithm (GA), which demonstrates improvement of model performance and enhances out-of-sample predictive accuracy. Our proposed model obtained a Mean Error (ME) of 16.66, a Root Mean Square Error (RMSE) of 66.84, a Mean Absolute Error (MAE) of 55.22, a Mean Percentage Error (MPE) of 0.30, and a Mean Absolute Percentage Error (MAPE) of 0.79. Our approach convincingly identifies linearity, seasonality, and nonlinearity imprints of the dengue progression in the USA. The model outperforms nineteen other techniques, including seven traditional, three Artificial Intelligence (AI) -based, a generative AI, and eight hybrid methods. The significance of the findings lies in a rigorous validation technique, specifically non-parametric tests, underscoring the practicability of the proposed model in dealing with noisy or incomplete data environments inherent in coping with health-related time-series challenges. Potential applications of the research could be the development of a robust early warning mechanism that empowers public health efforts with enhanced epidemiological surveillance.