Application of the Prophet Model for Predicting Explosive Dengue Outbreaks in Critical Regions of Peru: A Time Series Study, 2000–2024

Rojas-Paucar W.; Carranza López A.O.; Francia-Candiotti E.J.; Pareja-Granda E.A.; Medina Soto E.L. · Revista Cubana de Investigaciones Biomedicas (2025)

Resumen

Objective: To predict the progression of dengue cases in Peru, specifically in departments categorized as “explosive outbreak” areas, for the period 2025–2027 using the Prophet algorithm. Methods: Data from the Ministry of Health of Peru (MINSA) covering 2000–2024 were analyzed to develop forecasts. The Prophet model was trained with data from 2000 to 2023 and subsequently tested against real cases from 2024. Model performance was evaluated using RMSE, MAE, and MAPE error metrics. Forecasts were generated for 2025–2027, analyzing trends in each department. Results: The model performed well in departments such as Lambayeque, Callao, and Tumbes, but underestimated cases in La Libertad and Lima (errors > 16,000 cases). Conversely, in Madre de Dios and Ucayali, it showed persistent overestimations based on extremely high case values. Forecasts for 2025–2027 suggest a moderate annual increase of 10% to 20% in most departments, although caution is required in areas with high transmission rates. Conclusions: The model proved useful when cases followed stable trends but underperformed in sudden or extreme outbreaks, as it relies on univariate historical data. Therefore, predictions should be interpreted with caution, especially in departments where case numbers change abruptly.

Instituciones: Universidad Nacional Mayor de San Marcos; Universidad Nacional de Moquegua; Universidad Nacional Jorge Basadre Grohmann