Genetic algorithm for parameter estimation in double exponential smoothing

Abstract

The field of time series forecasting has grown up with the advent of greater computing power. During the past few decades, Genetic Algorithm (GA) has received a lot of attention. Due to its ease of applicability, numerous applications of GA are found. Double Exponential Smoothing are techniques that "smooths" the trend component in the data and are divided into Brown's One Parameter Linear Method and Holt's Two Parameter Method. One of the weaknesses in Double Exponential Smoothing methods is the parameters selection in model. This paper presents the development of a search procedure for parameter estimation in the Double Exponential Smoothing method using GA. The GA provides an alternative in determining the parameters in the Double Exponential Smoothing. Data used in this study are the daily Kuala Lumpur Composite Index (KLCI) and the daily USD/Ringgit exchange rate selling price in Foreign Exchange. A computerized system known as "DES system" was developed with the element of an interactive forecasting using the Double Exponential Smoothing method with the exploitation of GA. This software was written in Microsoft Visual C++ (with Microsoft Foundation Class (MFC). The result shows that the Double Exponential Smoothing using GA in searching for the parameter has greatly improved the forecast accuracy.