A Machine Learning Tool with An Integrated Dataset Towards the Construction of An Early Warning System for Dengue in Zulia State, Venezuela

• Autores: Cabrera, Maritza, Naranjo-Torres, José, Cabrera, Ángel, Zambrano, Lysien, Rodriguez-Morales, Alfonso J.
• Formato: artículo
• Fecha de Publicación: 2025
• Institución: Cuerpo Médico Hospital Nacional Almanzor Aguinaga Asenjo
• Repositorio: Revista del Cuerpo Médico Hospital Nacional Almanzor Aguinaga Asenjo
• Lenguaje: español
• OAI Identifier: oai:cmhnaaa_ojs_cmhnaaa.cmhnaaa.org.pe:article/2815
• Enlace del recurso: https://cmhnaaa.org.pe/ojs/index.php/rcmhnaaa/article/view/2815
• Nivel de acceso: acceso abierto
• Materia: Dengue fever; Machine Learning; Epidemiology; Disease Outbreaks; Fiebre de Dengue; Aprendizaje automático; Epidemiologia, Laboratorio, Hemodiálisis; Brotes de enfermedades

Introduction: The present study arises in response to the sustained increase in dengue outbreaks in Latin America, with special emphasis on the state of Zulia, Venezuela. This region, composed of 21 municipalities, is highly vulnerable to dengue transmission. Given this scenario, it is essential to have tools that allow early detection of outbreaks and, thus, optimize prevention and public health intervention strategies. The main objective is to develop an early warning system for dengue outbreaks using machine learning (ML) techniques. Materials and methods: Several data sources are integrated: epidemiological information, meteorological parameters, El Niño and La Niña (Niño 3.4 Index), socioeconomic and demographic variables. Two ML models were used: Support Vector Machine for regression (SVM-R) and Gaussian Process Regression (GPR). Results: The predictions obtained showed remarkable agreement with the actual dates on which the outbreaks were recorded, warning of the onset of dengue 2 to 3 weeks in advance, depending on the locality. However, in certain municipalities the predictions were less accurate, a finding that agrees with previous studies. Conclusions: In conclusion, the integration of epidemiological, climatological, and socioeconomic variables using ML techniques is presented as a promising tool for establishing.