Predicting crop yield is a critical challenge in modern agriculture due to the complex interplay of environmental, soil, and climatic variables. Traditional empirical methods have become increasingly unreliable as rapid environmental changes unfold. This paper presents a machine learning-based framework for crop yield prediction employing the Support Vector Machine (SVM) algorithm as the primary classifier. The study incorporates advanced data preprocessing, including feature selection methods such as Recursive Feature Elimination (RFE) and Modified Recursive Feature Elimination (MRFE), alongside data-balancing techniques including SMOTE and ROSE to address class imbalance. Comparative analysis against Naïve Bayes and Decision Tree baselines demonstrates that SVM achieves superior predictive accuracy. A Flask-based web application was developed to deliver real-time predictions from environmental input parameters. Results confirm that the proposed approach significantly improves prediction reliability, offering a scalable decision-support tool for farmers navigating volatile climatic conditions.

Keywords: Crop yield prediction; Support Vector Machine; Feature selection; SMOTE; Precision agriculture;