J. Porkodi, B. Karunai Selvi and A. Nagavaratharajan
Abstract
Fertilizers are indispensable for modern agriculture, providing vital nutrients crucial for crop growth, yield, and nutritional quality. They optimize soil fertility, alleviate nutrient deficiencies, and promote sustainable farming practices. This study explores the prediction of crop fertilizer needs using machine learning algorithms. The dataset consist of the following features such as Farmer Name, Father_or_Husband Name, Survey Number, Block Village, PH, EC, Organic Carbon (%), Nitrogen (Kg/ha), Phosphorous (Kg/ha), Potassium (kg/ha), Sulphur (ppm), Zinc (ppm), Boron ( ppm), Iron (ppm), Manganese (ppm), Copper (ppm) and Lime Status. These are the independent variables and the fertilizer is the dependent variable to be predicted. By assessing farmer details, soil attributes, and nutrient levels, Logistic Regression emerges as the most precise model to predict which major nutrient is required for the crop. Logistic Regression Algorithm, boasting an impressive 97% accuracy rate of prediction. Through meticulous analysis of classification reports, Logistic Regression proves superior among the algorithms considered, surpassing XG Boost, Decision Tree, and Support Vector Machine, which achieved accuracies of 92.7%, 90.5%, and 94.2% respectively. Hence machine learning technique proves to be an effective method for prediction of soil nutrient and as well as its analysis. This finding holds immense significance for agricultural practices. Precise fertilizer prediction optimizes resource allocation, boosts crop yields, and minimizes environmental harm. Harnessing machine learning empowers farmers to make informed decisions, tailoring fertilizer application to individual crop and soil requirements.