Vol. 10, Issue 10 (2021)

Author(s):
Mohanapriya M, R Vasanthi, Patil Santosh Ganapati and D Muruganandhi

Abstract:
This study involves comparing the predictive performance of the ARIMA model and ANN model. Monthly data of open price, close price, mean cash price, open interests and trading volume of maize futures are used for analysis. Trading volume is taken as dependent variable and all other variables are taken as independent variables. Data are checked for stationarity using Augmented Dickey-Fuller test based on the p-value. Trading volume and open interest are stationary at level and all other variables are stationary at first difference. The optimal ARIMA model was selected based on ACF plot, PACF plot and lower AIC, BIC values. The optimal model was found to be ARIMA (3, 1, 0). For ANN data set are divided into train and test set in the ratio of 7:3. Weights and bias are obtained for the built model. The forecast is obtained based on the coefficients of the model. The forecasted values are compared using RMSE, MAPE and MAE. Based on the results, ANN is better in forecasting the futures trading volume of maize than ARIMA.