Predição de casos de dengue na cidade de Fortaleza-CE utilizando internet das coisas e aprendizado de máquina

Abstract

Useful information from raw data generated by Internet of things devices can impact decision-making in public or private institutions. In healthcare, for example, one can predict epidemics or outbreaks. This work presents a contribution in the area of e-health more specifically within the scope of smart cities, using Internet of things and application of machine learning for dengue cases in weeks in the future. The city of Fortaleza-Ce was used as a case study in this work. It designs and implements an architecture composed of a weather data simulator, a back-end application and an integration with a IoT platform called dojot. A dojot was developed by the Research and Development Center in Telecommunications with the objective of developing technologies for smart cities, adapting to the objective of the work in question. The prediction of arbovirus cases is based on data obtained from the National Institute of Meteorology and the registration of dengue cases acquired from the Information System of Notifiable Diseases, of the Ministry of Health. A study carried out used the Coefficient of determination of Pearson and what was determined from Spearman to be correlated to the target. From this selection of data, a comparison of Machine Learning models was made using the MAE and the determination criterion R2 as a metric. After comparing the future, the model that presented the best levels of assertiveness was used to predict 5th week dengue cases in the future. In this way, the application case predicted by the architecture’s back-end is sent back to the dojot platform in the form of a notification.