Time series modelling for forecasting of food grain production and productivity of India

Abstract

The present study aimed for forecasting of total food grain production and productivity from 2018-2019 to 2025-2026 based on past history from 1950-51 to 2017-2018. Time series modelling and related forecasting were performed using Auto Regressive Integrated Moving Average (ARIMA), Auto Regressive Neural Network (ARNN) and ARIMA-ARNN hybrid models. ARIMA (0, 1, 1) were found suitable for the production and yield data based on the least value of Schwarz-Bayesian Criterion (SBC). Secondly, Auto Regressive Neural Network (ARNN) of order ARNN (3, 4) and ARNN (4, 3) was selected for both the dataset respectively. Lastly, ARIMA (0, 1, 1) - ARNN (3, 3) and ARIMA (0, 1, 1) - ARNN (3, 6) were found suitable for both production and yield. All the three models were tested for their forecast accuracy using Root Mean Square Error (RMSE) and Mean Absolute Percentage Error (MAPE). Accordingly, the ARNN model was found to be best as compared to the individual ARIMA and hybrid ARIMA-ARNN model. Based on the ARNN model, the forecasting of total food grain production and productivity calculated which would be 356.95 million tonnes with yield of 3183.67 kg/ha by 2025-26.