Análisis estadístico para pronósticos meteorológicos en un aplicativo web en la provincia de Chimborazo

Abstract

The meteorological variables of the stations issued by the Alternative Energy and Environment Group of the Escuela Superior Politécnica de Chimborazo are difficult to understand since it presents a constant climatic variation which prevents us from treating information with a defined pattern, therefore its objective is to determine forecasts through the use of three modeling methodologies defined by: Box-Jenkins, Neural Networks and Chaos Theory in order to help in making future decisions. The implemented methodology had a quantitative approach, a non-experimental design of a cross-sectional type was used since there was no direct participation in the manipulation of the variables in their natural environment, with a level of exploratory depth, an inductive study method because it was going to model meteorological variables and it was developed in a determined time. An exploratory analysis of the data was done in which it was obtained, as a general percentage, that 8.66% of the information are missing values in which the imputation of the data was performed through linear regression. In general terms, statistical techniques were applied to calculate forecasts. It is concluded that the best forecast method proves to be the Neural Networks, followed by the ARIMA models and then the Chaos Theory according to the Theil’s U coefficient and the Diebold- Mariano Test and the results of the meteorological forecasts according to stations will be reflected in a web application.