Vol. 19 No. 1
Evaluating forecast performance of GARCH model on weekly price of onion
Abstract: The development of an effective and trust worthy forecasting method for commodities with variable price series is essential in a country like India that is heavily dependent on agriculture. Due to the simultaneous presence of non-linearity, seasonality and complexity in the data, accepting a particular model for accurately forecasting price series of commodities like onion is difficult. In this endeavour, the performance of the time series model GARCH on the volatile weekly price series of onion of Kolhapur market of Maharashtra has been evaluated. To determine whether the series is stationary, Phillips Perron (PP) and Augmented Dickey-Fuller (ADF) tests have been applied. The Lagrange multiplier test was necessary to find the presence of the autoregressive conditional heteroscedastic (ARCH) effect. Values have been predicted for the next twelve horizons after the model was tuned with the training data set and the forecasts have then been compared with the testing dataset. Root Mean Square Error (RMSE) and Mean Absolute Percentage Error (MAPE) values have been used to determine accuracy. As seen, GARCH model has outperformed ARIMA model in dealing with the price dataset used in our study.
Download Full Article: Click Here