An Integrated AI Based Smart Crop Advisory System for Holistic Agricultural Decision Making in Maharashtra

Abstract

Agricultural productivity in Maharashtra is constrained by fragmented advisory services, limited access to real time information, and the absence of integrated decision support systems for smallholder farmers. This paper presents an AI based Smart Crop Advisory System that implements a unified and multi modal decision support framework by integrating machine learning, deep learning, and rule-based reasoning. The system comprises seven interconnected components, namely crop recommendation, yield prediction, fertilizer advisory, pest risk estimation, market driven profit prediction, image-based crop disease detection for cotton, maize, and tomato, and treatment recommendation. Real time weather and agricultural market data are incorporated through external application programming interfaces (APIs) to enable dynamic and context aware predictions. The system employs a random forest classifier for crop recommendation, regression-based models for yield prediction, and EfficientNet based deep learning architectures for disease classification. Rule based inference mechanisms and structured expert knowledge are utilized to generate fertilizer and treatment advisories. The framework is implemented as a lightweight web-based application, ensuring accessibility and usability for farmers and users with minimal technical expertise. Experimental observations indicate that the integrated advisory system effectively supports data driven agricultural decision making by combining predictive analytics, real time environmental inputs, and expert driven recommendations. The system is scalable, regionally adaptable, and demonstrates strong potential for enhancing sustainable agricultural practices and improving productivity in diverse agro climatic conditions.