Setting the Initial Value for Single Exponential Smoothing and the Value of the Smoothing Constant for Forecasting Using Solver in Microsoft Excel

Abstract

Although single exponential smoothing is a popular forecasting method for a wide range of applications involving stationary time series data, consistent rules about choosing the initial value and determining the value for the smoothing constant (α) are still required, because they directly impact the forecast accuracy. The purpose of this study is to mitigate these shortcomings. First, a new method for setting the initial value by weighting is derived, and its performance is compared with two other traditional methods. Second, the optimal (α) was automatically solved using Solver in Microsoft Excel, after which α was determined by minimizing the mean squared error (MSE). This was accomplished by comparing the α from Solver with step search by setting the smoothing constant by varying its value from 0.001 to 1 in increments of 0.001 and then choosing the optimal α value from this range that has the lowest MSE. The experimental results show that α from Solver and the optimal α with step search are not different, and the initial value set by the proposed method outperformed the existing ones regarding the MSE.