Determination of Demand Forecasting Variables by Exponential Smoothing Method

Abstract

Today, as in many sectors, many methods are used in demand forecasting, which is important in determining the most accurate strategies for production planning in the commercial refrigerator sector. The first of these methods is time series approaches. In particular, future demand trends of customers can be affected by many factors, such as market conditions and seasonality. It is important that the enterprises are planning the cost and product related to production, and that the resources are organized and structured in accordance with these goals. The purpose of these goals is to carry businesses to a solid place in the market financially and commercially by making accurate and reliable forecasts. In this study, the future estimations of dollar, TUFE, UFE and consumer confidence index variables, which will be used in the estimation of refrigerator sales in the production sector in Manisa between 2003 and 2022, were made using exponential correction (ETS) methods.

For the estimation of the other variables in the model to be established in the demand variable, the exponential correction method (ETS) was used and three different elements such as error, trend and seasonality were taken into account and estimation was made. The mean absolute percent error (MAPE) value was used to achieve the best performance among the ETS models. ETS models established with the available data are ETS (A, N, N) for TGE and ETS (M ,Ad, N) for TUFE, UFE and dollar. There is no seasonality in all these variables. In addition, while there is no trend in TGE, the effect of the trend decreases over time in other variables. However, with the flowing data, the data pool expands and the another ETS model can be found by using ETS methods. One of the biggest advantages of this method is that it allows new models by taking into account the flowing data easily.