Improvement of the Holt-Winters Multiplicative Method with a New Initial Value Settings Method

Abstract

The Holt-Winters method is a well-known and effective approach for forecasting time series, particularly when trends and seasonality exist. The proper settings of the initial values for level, trend, and seasonality play an important role in this method and lead to better forecasting results. In this paper, a new method is proposed to obtain the initial values in the Holt-Winters multiplicative method via setting the initial values of level and trend based on the weighted moving average and the initial value of seasonality based on a decomposition method. 10 real-world datasets were used to evaluate the performance of the proposed method compared to the existing Holt-Winters and Hansun’s methods while varying the smoothing parameter from 0.1 to 1 in increments of 0.1. The results of the study show that the proposed method outperformed the existing ones in terms of the mean-absolute-percentage error (MAPE), symmetric mean-absolute-percentage error (sMAPE), root-mean-squared error (RMSE), and the Theil-U statistic.