Small and marginal farmers in India often struggle with overproduction or underproduction due to unpredictable market demand and lack of precise yield allocation strategies. This study presents an ML-powered Crop Recommendation System that addresses these challenges by analyzing market demand and distributing crop yield requirements among farmers equitably. The system integrates three machine learning models: a composite model using SARIMA, Random Forest, XGBoost, ARIMA, and Prophet to predict yearly production requirements for major crops; a Random Forest Regressor to estimate crop yield based on land area; and a supplementary model that incorporates environmental factors for refinement. By focusing on market demand analysis, the system ensures that predicted crop requirements are proportionally allocated among farmers, reducing the risk of overproduction and waste. With high accuracy across its models, the system provides data-driven recommendations that align with market trends, empowering farmers with actionable insights to optimize crop selection and improve profitability.