Bayesian-Based Weather Prediction and Automated Distribution Simulation for Rice Harvest Optimization

Abstract

Weather variability creates substantial uncertainty in rice production and distribution, often resulting in discrepancies between predicted harvest quantities and logistical capacity. To address this challenge, this study presents an integrated framework that combines Bayesian weather prediction, climate-based rice yield estimation, and automated distribution simulation to support more reliable and uncertainty-aware agricultural planning. The Bayesian model is employed to generate probabilistic forecasts for key climatic variables rainfall, temperature, and humidity by producing posterior distributions that capture the inherent variability of environmental conditions, yielding stable predictive performance with an RMSE of 0.84 for rainfall and 0.62 for temperature. These probabilistic forecasts are subsequently utilized within a regression-based yield estimation model to quantify the influence of climatic fluctuations on harvest output, resulting in a mean absolute percentage error (MAPE) of 6.7%, which demonstrates strong consistency with actual production data. The estimated yields are then incorporated into an automated distribution simulation constructed as a weighted directed graph, where Dijkstra’s algorithm is applied to determine optimal delivery routes by evaluating distance, predicted load, and weather-related uncertainty. Simulation results reveal improvements in route efficiency and reduced deviation in travel times across varying climatic scenarios. Overall, the integration of Bayesian inference, yield prediction, and automated routing forms an adaptive and robust decision-support system for rice distribution management, offering a more reliable approach to optimizing agricultural logistics in environments characterized by dynamic and unpredictable weather patterns.