Comparative performance analysis of ARIMA and ANN models in rice production forecasting

Precision in projecting rice area, production, and yield is essential for effective agricultural planning, food security, and sustainable resource management, especially in structurally changing and nonlinear agricultural systems. This paper aims to conduct a comparative analysis of the forecasting capabilities of the Autoregressive Integrated Moving Average (ARIMA) model and Artificial Neural Networks (ANN) for annual rice area, production, and yield from 1990 to 2021. The study applies the Box–Jenkins ARIMA model with stationarity testing and ACF–PACF diagnostics, as well as a feed-forward ANN with backpropagation after appropriate data preprocessing. Mean Absolute Error (MAE), Root Mean Square Error (RMSE), and Mean Absolute Percentage Error (MAPE) are used to evaluate forecasting accuracy. The findings indicate that ANN consistently outperforms ARIMA across all three indicators, with lower forecast errors and greater stability, particularly in capturing the nonlinear dynamics of rice area and production. Although ARIMA performs reasonably well in forecasting yield, it demonstrates lower predictive power for area and production. Overall, the results confirm that ANN is a stronger and more reliable forecasting model for time-series analysis in the agricultural sector and can be of substantial value to policymakers and planners. Future studies may further improve forecasting accuracy by developing hybrid models, applying deep learning techniques, and integrating climatic or remote-sensing data.