Improvement for exponential smoothing

Exponential smoothing methods have been employed since 1950s and they aremost popular and used methods in business and industry for forecasting. Howeverthere are two main problems about choosing the smoothing constant and startingvalue. In this thesis a new method is introduced for smoothing constant and startingvalue. Modified method gives even more weights than the classical method to mostrecent observations. A software tool developed to compare the modified method withthe original. And real time series from M-competition are used to compare themethods empirically.