This work presents a complete crop recommendation system using a new ‘Blended XGSVM’ model, which integrates XGBoost and Support Vector Machine (SVM) algorithms to attain improved prediction accuracy. Source from the Karnataka state agriculture department and the “Raitamitra” portal, the proposed model is meant to examine agricultural data including soil, climate, and yield information. By means of careful preprocessing and exploratory data analysis (EDA), the system guarantees high-quality inputs for model training. Outstanding accuracy of 99.53% and sensitivity of 99.6% let the Blended XGSVM model surpass conventional techniques. The model’s computational complexity and the great tuning needed present difficulties even if it is quite effective. Aiming to give exact and dependable crop recommendations for farmers, future research will concentrate on optimizing the model for enhanced performance and scalability, so improving agricultural productivity.