Forecasting Model Accuracy Assessment in a Bottled Water Supply Chain

Abstract

This study presents an accuracy assessment of various classical time series forecasting methods to determine the most accurate forecasting method for predicting demand of the 50cl product of a bottled water supply chain. The classical time series forecasting methods compared are the moving average, weighted moving average, exponential smoothing, adjusted exponential smoothing, linear trend line, Holt’s model, and Winter’s model. The Mean Absolute Percent Deviation (MAPD) value was determined for the various forecasting methods to find the forecasting method with the least MAPD and hence the highest forecasting accuracy. The tracking signal measure was also used to determine the forecasting models which were biased. The results showed that though the weighted moving average method began to underforecast demand in the final period, it had the lowest MAPD value of 2.25%, making it the forecasting method with the highest accuracy for predicting the 50cl bottled water demand. On the other hand, though the exponential smoothing forecasting method had the highest MAPD value of 3.43% and least accuracy for predicting the bottled water demand, its forecasts were consistently within the tracking signal control limits. This study will aid supply chain analysts in implementing accuracy assessment of classical time series forecasting models.