Non-stationary model for rice prices in Bandung, Indonesia

A non-stationary forecasting model for rice prices is proposed in this paper. A daily rice price has two behaviors i.e. normal and spiky. To encounter both behaviors, non-stationary model has been applied. From various type of non-stationary model, the simple one which contains ARIMA model and GARCH model are applied. The ARIMA model is applied to catch the linear relationship of non-stationary data through differencing. On the other hand, the GARCH model is applied to unveil the heteroscedastic character of residuals with inconstant variance, a generalization of ARCH models. However, due to its limitation of recent GARCH model, MATLAB software and NUMXL, additional software on Excel, shall be used for parameters estimation of identified models. Hereafter, the forecasting of normal and spike patterns of rice prices are generated to form an overall price up to one day-ahead. As a result, this forecasting of the rice price could provide extensive and valuable information for rice market stakeholders.