Trade Volume Prediction Using Exponential Smoothing

Accurately forecasting financial indicators is challenging given their volatile nature, dependence on various market-related factors and impact on supply chains. Therefore, finding predictive models striking the balance between accuracy, simplicity of implementation and cost-effectiveness is of critical importance. In light of this context, many studies were conducted to design and evaluate high-performing models predicting financial indicators. This article is a preliminary study of the performance of parametric models in predicting volatile-prone time series. The adopted forecasting process enables the proper Exponential Smoothing (ETS or ES) model selection using three R functions, two horizon sets and different accuracy metrics. Since our data exhibits trend and seasonal patterns, we used the Holt-Winters (HW) Exponential Smoothing method to forecast the average monthly trading volume in a retail supply chain. The findings show that forecasting for shorter spans while updating the fit model selection (model building, choice of metrics) is recommended for better accuracy.