Intelligent crop recommendation system using supervised learning approaches

Abstract

This study presents a machine learning-based crop recommendation system designed to support precision agriculture in India. A dataset comprising 2,200 observations across 22 crops was used, integrating soil pH, fertilizer inputs, and climate parameters like temperature, rainfall, and humidity. Four classification models—Decision Tree (DT), Support Vector Machine (SVM), Random Forest (RF), and Artificial Neural Network (ANN)—were evaluated using accuracy, precision, recall, and F1-score. Random Forest achieved the highest accuracy (99.55%), followed by ANN (99.09%), SVM (98.64%), and DT (98.18%). While RF demonstrated the highest generalization and stability, ANN effectively captured complex, nonlinear relationships. DT was noted for its ease of interpretability. The results highlight the potential of supervised learning models to enhance site-specific crop selection and informed farm-level decision-making. The study also recommends future integration with real-time weather APIs, remote sensing, and mobile interfaces to expand reach and usability. This research contributes toward sustainable agriculture, with practical implications for smallholder farmers and climate-resilient crop planning across India’s varied agro-ecological zones.