Design of an Efficient Model for Satellite Image Classification Using Graph Neural Networks and Elephant Herding Optimization

This paper addresses the critical domain of Satellite Image Analysis, crucial for categorizing images into Forest, Water, and Urban Areas accurately. Existing techniques often yield suboptimal results due to complexities. Our novel approach combines Graph Neural Networks and Elephant Herding Optimization, showcasing significant advantages in precision (8.5%), accuracy (8.3%), recall (4.9%), processing speed (7.5%), and Area Under the Curve (AUC) (3.9%). This improvement stems from capturing spatial dependencies and fine-tuning model parameters effectively. Beyond Satellite Image Analysis, our work extends to land cover monitoring, environmental conservation, urban planning, and disaster management. The fusion of Graph Neural Networks and Elephant Herding Optimization highlights the potential of innovative deep learning techniques in remote sensing and geospatial analysis, aiding in a more sustainable decision-making process.