Uma Abordagem Baseada em Redes Neurais para a Predição de Chuva em Manaus, Amazonas


The weather in Manaus, Amazonas is typically Equatorial warm and humid which causes a natural discomfort for its inhabitants. One of the main agents for the local weather regulation is the precipitations, whose prediction is very important although difficult, because the weather in the city is under the influence of many precipitation systems. This work aims at presenting an appro- ach based on multilayered artificial neural networks to predict the occurrence of rainfall in Manaus. Data from an automatic meteorological station from 1970 to 2014 was used. After testing a set of 2500 different neural network that ad- dress this problem, it was possible to identify the network with 4 − 9 − 7 − 1 architecture with best performance, being able to predict rainfall with almost 70% certainty.

In: I Congresso Amazônico de Computação e Sistemas Inteligentes, Manaus, Amazonas.