Forecasting Commodity Prices with Exponential Smoothing
Name and surname of author:
Forecasting, Exponential smoothing, Commodity Exchange, London Metal Exchange, Industrial Metals.
DOI (& full text):
Analyzing and forecasting price changes belongs to very actual themes nowadays. In comparison with the past, much more attention and researches are dedicated to forecasting methods also at the…more
Analyzing and forecasting price changes belongs to very actual themes nowadays. In comparison with the past, much more attention and researches are dedicated to forecasting methods also at the universities. In the following text we will present possibility of forecasting time series of settlement prices of chosen metals at the London Metal Exchange (LME), using extrapolation of time series. More specifically, we will forecast monthly averages of cash seller and settlement price of chosen metals (Tin, Aluminum, Copper and Zink) using exponential smoothing method. This method belongs to classic mathematical / statistical methods of forecasting, named also quantitative methods - based on technical analysis. To examine the mentioned method, we performed an empirical experiment using real data of LME. We used average monthly cash seller and settlement prices of chosen metals, in period from January 1990 till June 2006. The effort was to forecast monthly averages of metal prices 6 months ahead, while forecasts were done always from the beginning of each quarter from January 2001 till June 2006. Data, from which the forecasting model of exponential smoothing was outgoing, was always from January 1990, till quarter from which the model carried out a forecast on next 6 months. Forecasts achieved by this process were compared with real prices from LME, available in time of experiment realization. For the success measurement of each forecast were calculated forecast errors - M.E. (mean error), M.S.E. (mean squared error), M.A.E. (mean absolute error), M.A.P.E. (mean absolute percentage error), M.P.E. (mean percentage error) and SSE (sum of squared error).