A time series analysis approach to forecast price fluctuation points

Abstract

ABSTRACT For manufacturers, material procurement and product sales directly influence cash flows. Buying materials at lower prices and selling products at higher prices is the easiest way to increase revenue. If a company can predict the price fluctuations in materials or products accurately, then it can maximize its profit by taking the appropriate action at the right time. This study proposes a novel price fluctuation forecast model, that is, the price fluctuation point forecast approach (PFPFA), to anticipate price fluctuation points and the degrees of price changes. Price fluctuation data are non-uniform time series. Hence, this study focuses on the quantities that predict price fluctuation points rather than the price fluctuation time epoch. The PFPFA encompasses four models to predict the price fluctuation points and three models to predict the price differences between periods. Thus, for each product, 12 forecast outcomes are possible. Applied to a real-world scenario and compared with exponential smoothing, the PFPFA offers acceptable predictions about price fluctuation points and performs better with regard to the degrees of price changes.