A Comparative Machine Learning Approach for Predicting Agricultural Productivity under Climate Variability

Abstract

Agriculture is highly dependent on climatic and environmental conditions, making agricultural productivity vulnerable to changes in temperature, rainfall, humidity, and other weather-related factors. Increasing climate variability creates difficulties for farmers and agricultural planners in estimating future crop productivity and managing resources efficiently. Conventional statistical approaches may have limitations when the relationship between climate variables and crop productivity is nonlinear and involves interactions among several factors. Machine learning provides an alternative approach for learning complex relationships from historical agricultural and climatic data. This research proposes a comparative machine learning framework for predicting agricultural productivity under climate variability using Multiple Linear Regression (MLR), Decision Tree Regression (DTR), Random Forest Regression (RFR), and Extreme Gradient Boosting (XGBoost) Regression. Climatic variables such as temperature, rainfall, humidity, and solar-related indicators, together with agricultural variables such as crop type, cultivated area, and historical yield, can be used as explanatory variables. Agricultural productivity is considered the target variable. The proposed framework consists of data collection, preprocessing, exploratory analysis, feature selection, model development, prediction, and performance evaluation. The models are evaluated using Mean Absolute Error (MAE), Mean Squared Error (MSE), Root Mean Squared Error (RMSE), and coefficient of determination (R²). The study is designed to determine which regression approach provides the most reliable prediction while also examining the relationship between climate variability and agricultural productivity. The proposed framework can support agricultural planning, climate adaptation, resource allocation, and data-driven decision-making.