Machine Learning Approach for Yield Estimation and Crop Prediction

Abstract - Optimal crop recommendations and yield prediction are essential for improving farm productivity and food security. Conventional approaches to the prediction of crop yields are usually limited by environmental variability. The present work proposes a strong machine learning-based framework aiming to support farmers in the process of choosing the best crops and predicting yield potential more accurately. The suggested system combines three supervised learning techniques—Decision Tree, Random Forest, and XGBoost—to examine critical environmental and soil parameters like nitrogen (N), phosphorus (P), potassium (K), soil pH, rain, humidity, and temperature. These attributes enable accurate crop recommendations under given environmental conditions. For yield prediction, past agricultural information, including parameters like state, district, crop type, season, and cultivated area, is utilized. Extensive model training and evaluation demonstrate that Random Forest performs exceptionally well compared to other approaches, with 96%. Accuracy ensures farmers are provided with the correct crop yields and recommendations. To ensure maximum accessibility, the solution is implemented as a web application using Flask, providing an easy-to-use interface for farmers with minimal technical knowledge to enter farm parameters and get actionable recommendations. This system not only facilitates better planning and resource optimization but also aids in reducing risks from uncertain weather patterns and volatile market conditions. Key Words: Crop Prediction, Yield Estimation, Random Forest, Flask Deployment, Precision Agriculture