author(s)
Hanumanthaiah R and Abhishek Singh

abstract
An attempt is made to compare the Linear and Non linear time series models.The Box Jenkins Auto regressive integrated moving average (ARIMA) and Generalized auto regressive conditional heteroscedastic (GARCH) models are studied and applied for modeling and forecasting of Prices and Arrivals of Black pepper in Bengaluru Market. Augmented Dickey Fuller (ADF) test is used for testing the stationarity of the series. ARCH-LM test is used for testing the volatility. It is found that ARIMA model cannot capture the volatility present in the data set whereas GARCH model has successfully captured the volatility. Root Mean square error (RMSE), Mean absolute error (MAE) and Mean absolute prediction error (MAPE) were computed. The GARCH (2, 1) and GARCH (1, 3) were found to be better models in forecasting prices and Arrivals of Black pepper in Bengaluru Market. The values for RMSE, MAE and MAPE obtained were smaller than those in ARIMA (1, 1, 1) and ARIMA (1 0 2) models respectively for Prices and Arrivals. The AIC and SIC values from GARCH models were smaller than that from ARIMA model. Therefore, it shows that GARCH is a better model than ARIMA for estimating Monthly Prices and Arrivals of Black Pepper in Bengaluru Market.