Explorando aprendizado de máquina com multivariáveis para previs ão de enchentes em ambientes IoTs: um estudo empírico no sistema de monitoramento de rios E-noé

• Lucas Augusto Vieira Brito
• Danielle Bressiani
• Jó Ueyama

Resumo

In the last years, WSNs are being implemented in different types of applications in the urban scenario, being one of the approaches, the monitoring of natural disasters, for example floods. Typically, applications to detect natural disasters are installed in inhospitable locations and rely on multihop communication for the data to reach a sink sink. In this scenario, one of the main challenges of these systems is to issue warnings in a timely manner to avoid major disasters. The issue of risk and disaster management and mitigation has been discussed by several countries at the conventions on climate change and sustainable development. However, correlating WSN disseminated data to achieve this goal is not a trivial task. For these reasons, this paper proposes a prediction model based on data mining and machine learning, in order to correlate different databases to achieve a higher quality of information. The algorithm Random Forest stood out among others, reaching 97% accuracy, and outperformed them in the rigorous t-test test of the weka tool. In addition, the forecast model has shown that it can better manage the resources of an RSSF.