Demand Forecasting of The Automobile Sales Using Least Square, Single Exponential Smoothing and Double Exponential Smoothing

Abstract

Sales strategies require the right managerial in marketing products with the development of technology and communication, the decision making in product sales supported by complete data and can be analyzed into intelligence business solutions. The research discussed and provided solutions about how to forecast future demand targets from a set of data history by making a predictive model of product demand in the real case. The research study was obtained from automobile sales, which the company probably set the strategy from the forecast result of automobile sales by the system in the future. The research used forecasting methods such as Least Square, Single Exponential Smoothing, and Double Exponential Smoothing to achieve a small percentage of prediction error. The dataset was collected from Mitsubishi Motors Corporation which obtained 60 samples of popular product types such as Pajero, FE and L300 from 2014 to 2018 over a period of months. The experimental results reported that Double Exponential Smoothing has given a better performance than other methods. The forecasting result of Pajero reached the MAPE of 3.26%, FE reached the MAPE of 3.24%, and L300 reached the MAPE of 3.37%. This study indicates that the selection of the forecasting method depends on the actual data pattern and the adjustment of the parameters in predicting future points.