AI-Powered Crop Yield Prediction and Optimization

Abstract

Agricultural productivity is heavily influenced by unpredictable climatic conditions, soil variability, and inconsistent farming practices, making yield estimation a persistent challenge for farmers and policymakers. Traditional yield forecasting relies on manual recordkeeping and expert judgment, which is time-consuming, subjective, and often inaccurate under changing environmental conditions. This work presents an AIpowered crop yield prediction and optimization system that leverages machine learning algorithms, namely Random Forest, Decision Tree, and KNearest Neighbors, to forecast crop yield based on soil nutrients, rainfall, temperature, and historical cultivation data. The system is deployed as a Flaskbased web application with an SQLite backend, offering an interactive interface for farmers to input field parameters and receive yield predictions along with optimization suggestions such as recommended crop choices and resource allocation. An administrative panel allows dataset management through CSV upload and on-demand model retraining. Comparative evaluation of the three algorithms shows that ensemble-based Random Forest achieves the highest prediction accuracy. The proposed system supports data-driven decision-making, helping farmers improve productivity while minimizing resource wastage.