An Optimized Hybrid ARIMA-LSTM Model for Time Series Forecasting of Agricultural Production in India

The role of agriculture in a country’s national income is very important. The zero-hunger sustainable goal of the United Nations cannot be achieved without substantial increase in agricultural production. The value of agricultural production shows the health of agriculture and its prediction has long been a challenge for academicians due to highly uncertain weather conditions as well as emergence of microbacteria due to variable soil and weather conditions. This chapter is focused on the prediction methodology for measuring accurate prediction of the value of agricultural production. The data of the gross value-added series for the years 1950–2021 for the agriculture and allied sectors has been taken for prediction. The traditional model autoregressive integrated moving average (ARIMA) and machine learning and the deep learning-based long short-term memory (LSTM) model have been compared with an optimized hybrid model comprising characteristics of both models. The efficiency of the models was checked by using root mean square error (RMSE). The result was compared with other studies on hybrid models. It has been reported that the hybrid model is more appropriate in comparison to both ARIMA and LSTM. The hybrid model provides the lowest RMSE among the three models. The ARIMA model gives lower RMSE in comparison to LSTM. The hybrid model provides 11% more accurate prediction in comparison to ARIMA. The prediction result of the hybrid model is highly significant and gives within the 95% confidence interval. The results are important and useful for the prediction of the value of agricultural production and policy formulation to meet the sustainable development goal of zero hunger.