Comparison of data mining models applied to a surface meteorological station

Abstract

This paper presents the application of data mining techniques for pattern identification obtained from the analysis of meteorological variables and their correlation with the occurrence of intense rainfall. The used data were collected between 2008 and 2012 by the surface meteorological station of the Polytechnic Institute of Rio de Janeiro State University, located in Nova Friburgo -RJ, Brazil. The main objective is the automatic prediction related to extreme precipitation events surrounding the meteorological station location one hour prior its occurrence. Classification models were developed based on decision trees and artificial neural networks. The steps of consistency analysis, treatment and data conversion, as well as the computational models used are described, and some metrics are compared in order to identify their effectiveness. The results obtained for the most accurate model presented a rate of 82. 9% of hits related to the prediction of rainfall equal to or greater than 10 mm h -1 one hour prior its occurrence. The results indicate the possibility of using this work to predict risk events in the study region.

Uso de sistema de controle fuzzy e internet das coisas para irrigação na agricultura de precisão

“…A irrigação apropriada e a sua consequente medição tem uma importância vital para a produção agrícola, pois culturas subirrigadas ou sobre irrigadas apresentam rendimentos de safras menores (CHARLES;NAMEN;RODRIGUES, 2017). Um modelo de irrigação que possa utilizar a previsão meteorológica de chuvas e desta forma reduzir o volume de água irrigada, e as suas consequentes perdas, associado a dispositivos de medição dos parâmetros do solo e do ambiente que tenham influência direta no desenvolvimento da cultura, poderá tornar o processo de irrigação agrícola mais eficiente, apresentando-se como uma solução ao problema do uso adequado dos recursos hídricos e também do aumento da demanda alimentar mundial.…”