Prediction of exchange rate peso-dollar using Artificial Neuronal Networks (ANR´s)

Abstract

The objective of the present work is to realize predictions of the type
of change peso-dollar being used Artificial Neuronal Networks (ANR´s),
for which, the investigation was based to determine the existing relation
between the obtained results and the effective types of change in the dates
of study, to determine the type of neuronal network that adapts more to the
prediction of types of change and to analyze the behavior of the variables
of the ANR in the process of prediction of the types of change. In order to
obtain this, using software Easy-N-extra, we selected information of twelve
economic variables of the year 2005 that served as entrance to a system of
neuronal networks, in that the exit was the type of change. Once realized
the training of the network and established the values of the variables of
entrance for the prediction process, the values of the type of change for
the first month of year 2006 were obtained; of this form, eighteen tests
were realized, using different combinations from variables. The obtained
results show to low allowable errors between the predictions and the real
results.