Paper Contents
Abstract
This research presents a machine learningbased crop yield prediction system designed to forecast agricultural production across Indian states using multi-parameter historical datasets. The study aggregates and preprocesses crop statistics, rainfall distribution, fertilizer usage, pesticide consumption, cultivated area and seasonal factors spanning multiple years. A Random Forest regression model, optimized through hyper-parameter tuning, is implemented as the primary predictor due to its strong handling of non-linear relationships within agricultural data. Model performance is compared against baseline regression techniques and ensemble variants, achieving high accuracy with R values ranging between 0.940.99, demonstrating strong generalization across major crops and regions. A user-interactive Streamlit interface is developed to allow real-time yield prediction based on variable input factors, supporting adoption at both local and policy levels. The proposed system has potential to assist farmers, agro-consultants and government bodies in planning cultivation strategies, managing resources and mitigating production risks. The research establishes a scalable framework for integration with satellite imagery and climate-forecast systems for enhanced future yield modelling.
Copyright
Copyright © 2025 Avaneesh Kumar, Aritra Pramanik, Chennuru Koushik Saai, K Komala Devi. This is an open access article distributed under the Creative Commons Attribution License.
