The accuracy comparison of time series model between Winters’ exponential smoothing and Box - Jenkins methods: A case study of forecasting garden coconut price

Abstract

          The objective of this research was to compare the accuracy of time series model using Winters’ exponential smoothing method and Box - Jenkins method for forecasting garden coconut price in Thailand. The set of data used in this research was taken from monthly garden coconut price during January, 2008 to December, 2016 (108 values). The first set of 96 values collected from coconut price during January, 2008 to December, 2015, was used to model time series model by Winters’ exponential smoothing method and Box - Jenkins method. The second set of 12 values collected from coconut price during January to December, 2016, was used to check the accuracy of the time series model via the criterion of the lowest mean absolute percentage error (MAPE). The results showed that;

          1)  The Winters’ exponential smoothing model was Y_t = (1327.57+102.89k) S_k+84 , when Y_t and S_k were forecasting value, and seasonal indices, respectively and the correlation coefficient of this model was 0.96.

          2)  The Box - Jenkins model was SARIMA (1, 1, 0), (0, 0, 0)₁₂ (no constant), Y_t = 1.51Y_t-1 - 0.51Y_t-1-2 + e_t when Y_t and e_t were forecasting and error value respectively, and the correlation coefficient of this model was 0.95.

          3) The Winters’ exponential smoothing model had MAPE = 40.83 and the Box - Jenkins model had MAPE = 41.93. The results showed that the Winters’ exponential smoothing model was a suitable for this data than Box - Jenkins model.