Stabilising market dynamics, directing agricultural policies, and guaranteeing food security all depend on accurate rice production predictions. The importance of creating accurate prediction models has grown due to the effects of urbanisation, policy changes, and climate variability.This study compares several time-series forecasting algorithms on the rice production data from 1961 to 2023, such as ARIMA, Holt's Linear Trend, and Simple Exponential Smoothing (SES), with improvements made utilising the Grey Wolf Optimiser (GWO) for optimal parameter selection. Furthermore, nonlinear dependencies in production patterns are taken into consideration by using the Generalised Least Deviation Method (GLDM). The findings show that the second-order GLDM model yields the most accurate forecasts for Bangladesh, whereas the GWO-ARIMA model performs best for India. Short-term autoregressive trends for China are well captured by the first-order GLDM model. The study emphasises the advantages of optimization-driven forecasting methodologies over traditional models, providing a strong basis for agricultural planning. These findings help policymakers and stakeholders implement data-driven methods for sustainable rice crop management.