Parameters estimation of holt-winter smoothing method using genetic algorithm

Abstract

A powerful technique based on adaptive heuristic namely Genetic Algorithm is widely used in many fields. This technique is very popular for solving global optimization problems. In this thesis, the Genetic Algorithm approach is used to estimate the parameters of Holt-Winter Exponential Smoothing method. The value of a combination of three parameters to be optimized, namely �, � and must lie between 0 and 1 by minimizing the one-step ahead forecasting accuracy of Mean Absolute Percentage Error (MAPE). Moreover, the difference of the initialization method, population size and crossover probability were also used, so that the comparative study of minimum value of MAPE can be done. The overall results of the Genetic Algorithm are compared with the conventional methods. From this study, it was found that the genetic algorithm outperformed the conventional method by giving the lowest value of MAPE. Hence, this proved that the genetic algorithm is effective for estimating Holt-Winter parameters. The data used in this study are monthly data set for the total number of tourist arrivals to Langkawi from 2002 until 2011. This investigation is done using computer simulations programmed by Microsoft Visual Studio 2010.