Rice Yield Prediction of Rajshahi, Dhaka, Chattogram, and Khulna Districts Based on Climatic Variables: A Machine Learning Approach

This research is about predicting rice yield in four major districts in Bangladesh—Rajshahi, Dhaka, Chattogram, and Khulna based on climatic variables using machine learning techniques. Agriculture is a key contributor to Bangladesh’s economy, yet farmers face several challenges in adapting to fluctuating climate conditions. This research leverages historical climate data, temperature, precipitation, humidity, and wind speed to analyze and discuss their impact on rice production. All the required data is sourced from reputable institutions to ensure accuracy and reliability. To develop an effective and reliable prediction model, multiple machine learning models such as auto regressive integrated moving average (ARIMA), multiple linear regression (MLR), random forest regression (RFR), and K-nearest neighbor (KNN) regression are applied. Among them, the RFR demonstrates the highest predictive accuracy as indicated by its superior R2 value. Each model's performance is rigorously assessed using established evaluation metrics to ensure robustness and reliability. The results provide valuable insights into the connection between climate change and rice yield, offering data-driven support for agriculture policymakers, farmers, and environmental planners. By integrating advanced machine learning models, this study enhances predictive capabilities, enabling better decision-making for sustainable rice production, which will improve the condition of agriculture. Despite certain limitations, the research establishes a strong foundation for future advancements in climate-aware agriculture forecasting and contributes to more proven farming practices in Bangladesh.