Crop Planning, Recommendations and Yield Estimation

Abstract

Agricultural decision-making in India is largely guided by traditional experience rather than scientific data, often leading to inconsistent crop selection, reduced yield, and unpredictable profitability. To address these challenges, this work presents a comprehensive web-based intelligent crop planning system that integrates machine learning, historical yield analytics, and scientific crop cycle knowledge to support data-driven agriculture. The system is designed with four functional modules. The first module provides crop recommendations by analyzing soil nutrient levels (NPK), pH value, rainfall, and real-time weather data obtained through an external API, using a Random Forest classifier for accurate prediction when trained model data is available. The second module generates a 12-week crop cycle plan that enhances soil fertility, prevents nutrient depletion, and minimizes pest recurrence through scientifically guided rotation strategies. The third module processes district-level historical yield datasets to identify region-specific high-performing crops, enabling better planning across diverse agro-climatic zones. The fourth module assesses the economic feasibility of cultivation by estimating revenue, cost, and profit per hectare using realistic reference values and user inputs. The system is implemented using React for the frontend, Flask for backend processing, and SQLite for managing user data. Experimental results demonstrate that lightweight machine learning models combined with carefully curated datasets can deliver reliable, interpretable, and actionable crop insights for farmers, agricultural planners, and researchers, thereby supporting sustainable and informed agricultural decision-making.