CROP YIELD PREDICTION USING MACHINE LEARNING TECHNIQUES

In recent years, the rise of big data has accelerated the adoption of machine learning approaches across
diverse domains. Agriculture, in particular, stands to benefit significantly from these advancements, with
the potential to enhance crop productivity and quality. This study presents a machine learning–based
framework for crop recommendation and yield prediction to support data-driven agricultural decisionmaking. Exploratory Data Analysis (EDA) is employed to uncover patterns in the dataset, with correlation
maps used to visualize feature relationships. The dataset is partitioned into training and testing subsets,
and preprocessing includes feature scaling via Standard Scaler and categorical encoding with
OneHotEncoder. The proposed architecture integrates Decision Tree Regressor (DTR), Support Vector
Regressor (SVR), and Gradient Boosting Regressor as base models, while XGBoost (XGB) and Random
Forest (RF) Regressor serve as meta-models within a Stacking Regressor pipeline. Stacking is selected over
bagging and boosting due to its ability to combine heterogeneous models in a meta-learning framework.
Model performance is evaluated using Mean Absolute Error (MAE), Mean Squared Error (MSE), Root Mean
Squared Error (RMSE), and R-squared (R²) metrics. Results demonstrate that the stacked ensemble
provides an effective and practical tool for enhancing agricultural decision-making through advanced
machine learning methodologies.