Hybrid Artificial Intelligence Framework for Soil Analysis and Crop Recommendation in Smart Agriculture

Abstract

Sustainable agricultural strategies that optimise food production while minimising resource depletion are in high demand due to the increasing global population. Agricultural output and decision-making are both enhanced by Smart Farming's crop suggestion and soil analysis, which take into account climatic conditions, environmental considerations, and soil fertility when deciding which crops to plant. This research presents a smart system that uses topography, soil characteristics, climatic variables, and past crop yield data to combine soil analysis with sophisticated machine learning. By standardising, addressing missing values, and encoding categories, data quality and consistency are enhanced. To find the most important traits and get rid of duplicates, a multi-stage feature selection method is used, which includes PCA, IG, and correlation analysis. After that, a hybrid model called Kernel ELM-Attn is used to forecast crop yield. The ST-GETN model is then optimised for accurate crop recommendation using the PantheraCobra metaheuristic. The experimental results demonstrate that the suggested framework outperforms existing methods and proposes crops with a 97.26% accuracy rate. The suggested method improves prediction accuracy, precision farming, soil-based decision-making, and long-term Smart Farming viability via resource optimisation and increased crop yields.