Intelligent Machine Learning Framework for Precision Agriculture: AI-Driven Cluster Analytics and Economic Decision Support

Abstract

Integrated analytical approaches are needed to address the crop productivity variability, climatic variability, soil characteristics, agricultural inputs and market behaviour which are key components of precision agriculture. This study proposed an intelligent machine learning system that combines AI based cluster analytics, crop yield prediction and economic decision support. The merged agricultural dataset consists of 19,689 observations from 30 states in India, 55 crops and six crop seasons. The data was preprocessed with geographical standardization, handling missing values, assessing outlier detection, encoding categorical data, and scaling numerical values. Feature engineering yielded indicators such as the crop yield, deviation from rainfall, soil nutrients, variation of market prices and estimated gross revenue. The two agricultural production environments were identified by using k-means clustering where the differences between them were measured in terms of usable area, wind, precipitation, temperature, relative humidity, soil nutrients and income. Cropped yield prediction models were evaluated for their effectiveness; there were four models. Random forest had the highest accuracy with an R^2 of ~0.931, while histogram gradient boosting, extra trees, or linear regression did not. Economic analysis of price matched crops revealed that onion had generated highest mean gross revenue whereas potato had a relatively favourable return–risk profile due to its lower price variability. The results show that a combination of agricultural clustering, the use of a nonlinear yield function and market-based revenue evaluation can offer more context-specific recommendations than do approaches based on yield or price alone. The framework proposed here provides a feasible approach for data-based crop planning, productivity evaluation and economically-informed precision-agriculture decisions.