Aplicação do algoritmo Multi-Layer Perceptron (MLP) para serviço de predição de dados com aprendizado incremental

Abstract

The climate is always changing and it is very important to monitor climate variations, since many human activities depend on specific conditions to be successful. Over time, several works related to climatic variations have been developed. Many of these works involve IoT (Internet of Things), considering that the monitoring of climatic quantities often requires sensing networks. The Sensor Nodes that make up these networks often suffer from the problem of power supply, due to a great dependence on batteries for nodes that are in remote locations. Computational intelligence is a resource that can be used to solve different types of sensor network issues. Chacon (2021) developed a data prediction service using Machine Learning, aiming to solve the problem of energy consumption of batteries used in Sensor Nodes. The work developed aims to find a different approach to develop data prediction models used in Chacon (2021), which initially did not have such satisfactory results. The elaboration took place through a more detailed study of the databases and the use of algorithms to find better parameters. The experiments performed presented models and training strategies capable of predicting the air humidity data, but not the temperature.