Forecasting of area and production of sorghum in Maharashtra, India using ARIMA and ANN models

Abstract

Sorghum, a highly valued crop in India, has been experiencing a decline in production over the past two decades, despite India being the largest producer and consumer of millets. Forecasting expected area under cultivation and production of sorghum in advance can play a pivotal role in reversing this trend. This study made an attempt to find suitable model for sorghum area and production forecasting in Maharashtra, India. From 1966 to 2021, data on sorghum area and production were gathered and forecasted using ARIMA and ANN techniques. ARIMA (0, 1, 1) with drift is fitted best for the sorghum area and ANN (3:2S:1) best-fitted model for sorghum production. The model's accuracy was compared using RMSE, MAE and MAPE measures. The study found the ARIMA model effectively forecasts sorghum areas, while the Time delay neural network model effectively captures heterogeneity and complexity in production data.