An AI-Driven Smart Crop Recommendation and Advisory Framework

Abstract

In the face of escalating climate variability, market uncertainties, and resource constraints, farmers increasingly require intelligent decision-support systems to sustain agricultural productivity. This study presents an AI-driven smart crop recommendation and advisory framework that integrates supervised machine learning algorithms with natural language processing to support sustainable agriculture in diverse agro-climatic regions of India. The system utilizes a curated dataset comprising soil nutrient values (N, P, K), pH, rainfall, and temperature, combined with crop price data from Agmarknet, to provide accurate crop predictions and dynamic profit estimations. Among various models evaluated—Random Forest, Decision Tree, and K-Nearest Neighbors—the Random Forest Classifier demonstrated superior performance with an accuracy of 87% and the lowest RMSE of 14.57. The framework further includes an alternative crop suggestion module, which uses Euclidean distance to recommend viable substitutes based on soil and climate proximity. Additionally, a Gemini API-powered AI assistant delivers personalized, region-specific advice on crop care, pest management, and weather, enabling effective interaction in natural language for digitally underserved farmers. The proposed system not only enhances decision-making at the farm level but also contributes to improved resource management and income predictability. Its scalable, modular design opens pathways for future integration of satellite data, IoT-based soil sensing, and adaptive learning strategies. Overall, this work underscores the potential of AI in driving data-informed, inclusive, and resilient agricultural practices.