Autoregressive Integrated Moving Average (ARIMA) Model with Genetic Algorithm to Forecast the Chilli and Turmeric Productions in India

Abstract

Aims: India holds the distinction of being the foremost producer of spices globally and has been long-run history in spice export. The quantity of Indian spice exports increased by 37% with $ 4.1 billion worth in 2021. With that, dried chilli, cumin, and turmeric alone contributed 44% of export value ($ 1.8 billion). Forecasting the production of major spices are key for exports and plays an essential role in supporting and achieving the target of $10 billion in exports by 2027. Data Source: The time series data of chilli and turmeric production data in India from 1970-2020 periods was collected from Indiastat. Methodology: The present study sought to forecast the production of chilli and turmeric in India using the ARIMA model and their parameters are estimated by stochastic optimization techniques (genetic algorithm). The parameters are estimated by minimizing the Mean Absolute Percentage Error (MAPE). Finally, ARIMA and ARIMA_GA models were compared based on their predictive ability. Results: The Root Mean Square Error (RMSE) and Mean Absolute Percentage Error (MAPE) were 254.01,11.32 (chilli) and 185.73, 15.24 (turmeric) for testing set of ARIMA_GA model which is lower than the fitted ARIMA model. Conclusion: This work has shown that ARIMA_GA (2,1,1) has been the best model to forecast the chilli and turmeric production in India. ARIMA_GA model will cope with parsimony and convergence of likelihood function to global optimum problems. Therefore ARIMA with GA will able to model the complexity and uncertainty of the data.