Abstract:
Instability of prices of coconut and coconut products is a well-known phenomenon in Sri Lanka during the last decade. Therefore, an appropriate forecasting model of prices is needed for producers, consumers and policy makers. Objective of the study was to find an appropriate model to forecast coconut and allied products in Sri Lanka. Published data of coconut prices of Coconut Development Authority (CDA) and Central Bank of Sri Lanka (CBSL) for the period 1974 to 2004 were used for the analysis. Time series plots were used to find major behavioral patterns. If the data series were non-stationary, first order differences were taken. Further, first order differences of log-transformed series were taken, if the data series were non-stationary after having first differences. Stationary data series were fitted to six conventional time series models, viz; General decomposition method, moving average method, Winter’s method, Single exponential smoothing method, Double exponential smoothing method, and Auto regressive moving average (ARIMA) method. Mean Absolute Percentage Error (MAPE), Mean Absolute Deviation (MAD), Mean Squared Deviation (MSD) were used to test the model adequacy and accuracy. The results revealed that the ARIMA & exponential methods are better than other models to predict prices of coconut and coconut products in Sri Lanka