A Comparison of ANFIS, ANN and DBR systems on volatile Time Series Identification

This paper shows a comparative study for the US Dollar - Colombian Peso Exchange Rate identification case using statistical models like ARIMA, ARCH and computational intelligence techniques like ANFIS, Neural Networks and DBR. This study case is specially interesting because is a Time Series that presents a volatile behavior and complex problem for classical analysis. The technique selection method is based on statistical theory and tests, which are appropriately criteria for selecting an alternative. Time series statistical theory and methods are used to select an adequate technique, based on residual analysis and classical Time Series test for model adequation. Bayesian, Akaike and Swartchz criteria, Mc Leod-Li, Ljung-Box, ARCH, Turning Points and other randomness tests are used to select the best estimated option.