Use of regression model and ARIMA model for forecasting kharif food grain production of Odisha: A comparative study

Abstract

Present work discusses the issue related to the model selection for efficiently forecasting the area, yield and hence production of food grains grown in Odisha. Several models have been tried on the observed data on area and yield for the period from 1992-93 to 2010-11and the best model have been selected by comparing the model fit statistics after testing the model diagnostics criteria. The models tried are ordinary regression models, spline regression models and ARIMA models. The model diagnostic criteria used are Shapiro-Wilk’s Statistic and Durbin-Watson statistic. The model fit statistics used are R2 , adjusted R2 and Root Mean Square Error (RMSE). The selected models are also cross validated by using the known values for the year from 2011-12 to 2015-16. The cross validation of the selected model test the efficiency of the model in forecasting. Lastly, the best selected model has been used for forecasting area and yield of kharif food grains. Using the forecast values of area and yield the forecasts are obtained for production of kharif food grains.