Improving Monkeypox Outbreak Prediction Through Time-Series Forecasting with Machine Learning Models

The global resurgence of infectious diseases, notably Monkeypox, has raised significant public health concerns due to its rapid spread and unpredictable nature. Traditionally endemic to Central and West Africa, recent outbreaks in nonendemic countries emphasize the need for accurate forecasting tools. This study applies advanced machine learning models: XGBoost, Long Short-Term Memory (LSTM) networks, and Prophet to predict Monkeypox-related deaths using time series data from 2019 to 2023. The dataset includes new cases, new deaths, and time-based features like year, month, and week of the year. Feature engineering techniques, such as lag features and rolling statistics, were used to capture temporal and non-linear relationships in the data. The results show that Prophet outperformed other models, achieving the lowest mean square error (MSE) of 0.4175 and the highest R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> of 0.5911, demonstrating its ability to model seasonality and trends effectively. XGBoost, when enhanced with lag features, ranked second with an MSE of 0.4832 and R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> of 0.5503. LSTM performed moderately well with an MSE of 0.5567 and an R2 of 0.4935, although it struggled with peak seasons. This study highlights the effectiveness of machine learning models in providing accurate predictions for public health interventions. By integrating real-time data, these models can significantly improve the forecasting of Monkeypox outbreaks, aiding in better resource allocation and intervention strategies.