METHODOLOGY BASED ON MACHINE LEARNING AND DEEP LEARNING TO PREDICT DENGUE TRANSMISSIONS

Sosa Callupe S.A.; Rojas Vales N.C.; Taipe Ramos L.N.; Blas Morales O.E.; Palacios Bett J.S.; Eulogio Saenz M.; Quintana M. · Proceedings of the International Astronautical Congress, IAC (2022)

Resumen

The Sustainable Development Goals (SDGs) Sustainable Cities and Communities 11 and Climate Action 13 cover the problems of flooding and its consequences, such as the proliferation of diseases that threaten public health. This study correlates flooded areas, temperature, humidity and average precipitation parameters with the level of dengue contagion in the areas surrounding the Piura River in Peru through remote sensing, artificial intelligence and various data sources. The prediction model is a decoder recurrent neural network that receives as input parameters the flooded areas, temperature, humidity and precipitation for 20 weeks, and returns an estimate of the number of cases of people infected by dengue during the next two weeks. Public databases from state institutions were used to train the neural network. The parameters of temperature, humidity and precipitation averaged by month were obtained from the virtual platform of the National Meteorological and Hydrological Service of Peru (SENAMHI) and for the test we used the amount of wet area in the study area obtained through UNET and a multitemporal analysis of SAR Sentinel 1 images from the ASF-Nasa database. Dengue cases were obtained from the National Center for Epidemiology, Prevention and Disease Control (CDC-Peru). We achieved a prediction model of dengue infection cases for the district of Castilla; we complemented it with a SAR image processing program and manual segmenters; we only had as input the scene names of the Sentinel-1 multitemporal images by means of Colab. Finally, we can prevent the percentage of infections per month not only in the studied region, but also applied to areas with similar geographic and socioeconomic conditions; with an adequate prevention, we could avoid the health crisis that dengue can cause with the only condition that there is no lack of data for the study. Copyright

Instituciones: Universidad Nacional de Ingeniería