Time series prediction based on genetic algorithm with application in finance

Abstract

Real world problems are described by non-linear and chaotic processes, which makes them hard to model and predict. The aim of this paper is to determine the structure and weights of a time series model using genetic algorithm (GA). The paper first describes the traditional procedure of estimating time series models, which are commonly used in financial forecasting. These traditional estimation methods may not be adequate enough to capture stochastic nature of the financial time series due to its complexity. This article gives a brief background of Genetic algorithm method and its estimation procedure. This approach was then applied to model the Naira exchange rates against other currencies and it yielded a mean square error of 0.0058,0 .00799,0 .03711,1 .212 and 0.1108 for U.S dollars, British pound, Japanese Yen, CFA franc and Swiss franc respectively .