CROP RECOMMENDATION, YIELDPREDICTION&FERTILIZER PREDICTION (ML)

Abstract

The agricultural sector increasingly relies on data-driven solutions to optimize crop production, reduce losses, and enhance sustainability. Machine Learning (ML) has emerged as a powerful tool for analyzing soil parameters, weather conditions, and historical crop performance to enable intelligent decision-making. This paper presents an integrated ML framework capable of performing three key tasks: crop recommendation, yield prediction, and fertilizer prediction. The system utilizes environmental variables such as soil pH, nitrogen–phosphorus–potassium (NPK) levels, temperature, humidity, and rainfall. Multiple ML models including Random Forest, Support Vector Regression, and Gradient Boosting are trained using publicly available datasets. The proposed framework recommends the most suitable crop for a given farmland, predicts expected yield per hectare, and suggests optimal fertilizer composition to improve crop health and soil nutrient balance. Experimental results show that the integrated system achieves high accuracy, with Random Forest providing the best performance across the three tasks. The work demonstrates that ML-driven approaches can significantly assist farmers in making informed decisions, leading to improved productivity, reduced fertilizer overuse, and enhanced sustainability.