Open AccessGeoDisasterAINet: An Explainable Deep Ensemble Framework for Real-Time Urban and Rural Disaster Classification and Resilience
Author(s) -
Akella S Narasimha Raju,
Seelam Sreekanth,
Ranjit Kumar Gatla,
M. Rajababu,
Devineni Gireesh Kumar,
Aymen Flah,
Claude Ziad El-Bayeh,
Khaled A El-Nagdy,
Ali Alzaed
Publication year - 2025
Publication title -
ieee access
Language(s) - English
Resource type - Magazines
SCImago Journal Rank - 0.587
H-Index - 127
eISSN - 2169-3536
DOI - 10.1109/access.2025.3574451
Subject(s) - aerospace , bioengineering , communication, networking and broadcast technologies , components, circuits, devices and systems , computing and processing , engineered materials, dielectrics and plasmas , engineering profession , fields, waves and electromagnetics , general topics for engineers , geoscience , nuclear engineering , photonics and electrooptics , power, energy and industry applications , robotics and control systems , signal processing and analysis , transportation
These Sustainable Development Goal 11 (SDG 11), whose target is to make urban and rural settlements resilient and disaster-prepared, integrates with it the significant objective of having disaster-resilient communities and cities. GeoDisasterAINet is an AI deep learning and machine learning model for distinguishing between rural and urban-area occurrences of floods, cyclones, earthquakes, and wildfires effectively. In this article, we introduce GeoDisasterAINet. There are four successive phases in model performance improvement, with increasingly enhancing classification accuracy, namely an integrated CNN (baseline model), an integrated CNN with XGBoost (feature improvement model), an integrated CNN with XGBoost and Multiclass SVM (refinement model for decision boundary), and LIME segmentation for interpretability and model performance evaluation.EfficientNetB7, ResNet-50, InceptionV3, and DenseNet-201 are architectures in use in GeoDisasterAINet, and these architectures make three settings for an ensemble: ERI-2025 (EfficientNetB7, ResNet-50, InceptionV3), DRI-2025 (DenseNet-201, ResNet-50, InceptionV3), and DE-2025 (DenseNet-201, EfficientNetB7). In balancing out the dataset, we used the Synthetic Minority Over-sampling Technique (SMOTE). With 99.25% accuracy in training, 96.37% accuracy in testing, 98% in precision, 97% in recall, and 97% in F1-score, ERI-2025 + XGBoost + Multiclass SVM, our best model, effectively classified items. LIME-based interpretability uncovered disaster-specific features such as storm behaviour in cyclones, infrastructure loss in earthquakes, wildfire propagation in wildfires, and water build-up in floods, and in confirming that the model was reliable, it helped in an effective location of disasters. GeoDisasterAINet is a strong AI-powered system that addresses directly SDG 11 through its contribution towards safer and disaster-resilient communities and cities. It does this through AI-powered deep learning and machine learning technology for improving disaster detection, evaluation of risk, and preparedness for emergencies in real-time.