Traditional Machine Learning based on Atmospheric Conditions for Prediction of Dengue Presence

• Autores: Sánchez López, Brenda Sofía Zoila, Candioti Nolberto, Daniela, Taquía Gutiérrez, José Antonio, García López, Yván Jesús
• Formato: artículo
• Fecha de Publicación: 2023
• Institución: Universidad de Lima
• Repositorio: ULIMA-Institucional
• Lenguaje: inglés
• OAI Identifier: oai:repositorio.ulima.edu.pe:20.500.12724/19484
• Enlace del recurso: https://hdl.handle.net/20.500.12724/19484; https://doi.org/10.13053/CyS-27-3-4383
• Nivel de acceso: acceso abierto
• Materia: Machine learning; Forecasting; Aprendizaje automático; Dengue; Prospectiva; https://purl.org/pe-repo/ocde/ford#3.03.05

The dengue virus has become an increasingly critical problem for humanity due to its extensive spread. This is transmitted through a vector that sprouts in certain climatic conditions (tropical and subtropical climates). The transmission of the disease can be associated with certain climatic variables that reinforce the outbreak. Data were collected on dengue cases by epidemiological week registered in Loreto-Peru from January 1, 2016, to January 31, 2022. Likewise, data on meteorological variables (maximum and minimum temperature; dry and humid bulb temperature; wind speed and total precipitation in the area). In this study, four Machine learning modeling techniques were considered: Support Vector Machine (SVM), Decision Tree, Random Forest and AdaBoost; and the parameters defined to evaluate the models are: Accuracy, Precision, Recall and F-1. As a result, optimal AUC values were obtained in a range from 0.818 to 0.996 for the SVM, Random Forest and AdaBoost algorithms, likewise, in all weather stations the ROC curve showed good performance for all models, except for the Decision Tree algorithm. As a conclusion for this study the optimal model to associate dengue cases with climatic conditions is SVM.