HOLT’S EXPONENTIAL SMOOTHING AND ARIMA MODELS FOR FORECASTING COCONUT PRODUCTION TRENDS IN INDIA

Abstract

This study forecasts coconut production in India for the next decade using two time series forecasting models: Holt’s Exponential Smoothing and ARIMA. By analysing historical production data from 1956 to 2021, both models indicate a consistent upward trend in coconut production. The study compares model performance using accuracy measures such as MAE, MAPE, and AIC. Results show that while both models forecast similar trends, ARIMA  outperforms Holt’s Exponential Smoothing in terms of accuracy, as reflected by lower MAPE values. Although Holt’s method fitted the values slightly better, ARIMA produced more reliable predictions. The models did not perform well based on RMSE criteria, but residual analysis suggests that Holt’s method produces more random, white noise residuals compared to ARIMA. The findings underscore ARIMA’s superior predictive accuracy in forecasting coconut production, despite both models having potential for long-term forecasting.