Using autoregressive integrated moving average models to forecast the production and yield of rice in Uttar Pradesh

Abstract

The present study was conducted to forecast the production and yield of rice for U.P. grown during the period 1966-67 to 2020-21. Data were analysed by using time-series methods namely suitable ARIMA model. The entire data set has been splited into two parts in training data set and the testing data set form. The stationarity of the series has been checked and the presumptive p and q values are decided using the ACF and PACF plots. Next, using these presumptive values of p and q, two possible ARIMA models are selected, these are namely ARIMA (1,1,0) and ARIMA (1,1,1), in support of forecasting the yield and production of rice. L Jung-Box test and Shapiro-Wilk’s are used to test residual diagnostics and normality of errors respectively. Some model fit criteria viz; Akaike information criteria (AIC), Bayesian information criteria (BIC), Mean absolute percentage error (MAPE) and Root mean square error (RMSE) values are observed for selecting the best ARIMA model. Here, ARIMA (1,1,0) model has been found to be best and used for forecasting production and yield of rice. It is found that MAPE is 15.52 and 16.40 for model ARIMA (1,1,0) which is comparatively best in the class of all candidate models for forecasting production and yield of rice of UP, respectively. Using this best-fitted model, the targeted forecast values for production and rice yield for the upcoming five years starting from 2021-22 to 2025-26 are obtained.