Previsão de epidemias através do Twitter

Abstract

Aiming at better management of resources and democratization of health services, the DATASUS offers a database with information relevant to the quantification and evaluation of health information. However, one difficulty is that not always the data obtained through official channels are made available in a timely manner and often are only available when an epidemic is out of control, not giving time to preventive measures of public agencies. In this context there is a need to investigate other methods that allow the collection and analysis of data to facilitate the dissemination of information required for preventive actions in health. Therefore, the main objective of this dissertation is to propose a methodology for analysis of epidemics from the social network Twitter. To do this, we conducted a study in which we sought to identify suspected cases of Chikungunya fever in Brazil from the symptoms reported by users on the social network. Then, with the use of text mining techniques were checked 258,707 tweets with any symptoms of the disease during the period from 15 August 2015 to 31 January 2016. The results show a high correlation between the cases verified in the social network with the cases notified by the Secretariat of health surveillance, which indicates that the social network of Twitter can be used as a means for analysis and prediction of epidemics. In this way, we can also conclude that the symptoms can be used as a parameter for detection of outbreaks of the epidemic.