Abstract
Exponential smoothing (ETS) method and autoregressive integrated moving average (ARIMA) models have been
applied to forecasting monthly price of logs, sugi (Japanese cedar, Cryptomeria japonica D. Don) and hinoki (Japanese cypress, Chamaecyparis obtusa (Sieb. et Zucc.) Endl.). In this research, we evaluated the forecast accuracy of these two approaches through cross validation. Monthly current price data from January 2002 to December 2016 were used. The results show that the forecast accuracy of ARIMA models was not statistically significantly different from that of ETS method at 5% level for the period from January 2010 to December 2016 under all 12 forecast horizons of one month to 12 months. By comparing with the amount of changes in original prices, it was found that both ETS method and ARIMA models forecasted with smaller mean absolute errors than mean absolute amount of changes in prices within 8 months of forecast horizons at 5% level, which showing that ETS method and ARIMA models outperformed naive method. However, by comparing with seasonally adjusted naive method, ARIMA model showed better forecast accuracy only for forecast horizons of 1 and 2 months for hinoki at 5% level of statistical significance, while ARIMA model at other horizons and ETS method did not outperform seasonally adjusted naive method in forecast accuracy.
