Diseño e implementación de un algoritmo genético para la predicción de una variable

Abstract

This article describes the application of a genetic algorithm to obtain a better forecast precision of any variable using historical data. In order to measure this precision, the exchange rate Mexican Peso – US Dollar in Mexico was used in comparison with the predictions calculated by the statistical method moving averages. The problem consisted of obtaining the values that minimize the average quadratic error between the real value and the forecast value to obtain a prediction with a minimum error margin. The genetic algorithm application was designed using the following real representation chromosomes: the crossing BLX-0.5 operator as well as the non-uniform mutation operator. These operators offer a better capacity on exploration and exploitation resulting in the genetic algorithm that provides a precision increase of 14% in comparison to the precision of the statistical moving average method.