Big data analytics has become a cornerstone of modern intelligent systems, yet its complexity and opacity often limit interpretability, trust, and security in decision-making. To address these challenges, we present the Hybrid DeepXAI Fusion Network, a novel deep learning framework designed to fuse temporal, relational, and contextual features while embedding explainable artificial intelligence (XAI) modules for transparent inference. The system leverages 1D-CNNs for temporal feature extraction, Transformer encoders for long-term contextual modeling, and Graph Neural Networks (GNNs) for capturing inter-feature dependencies, integrated through an attention-based multimodal fusion mechanism. Additionally, Class Activation Mapping (CAM) and explanation loss are incorporated to ensure interpretability consistency, enhancing trustworthiness in sensitive applications. The model is deployed in a federated cloud architecture, ensuring data privacy through homomorphic encryption and blockchain auditing. Experimental evaluation across healthcare, finance, and energy datasets demonstrates superior performance, achieving 98.42% accuracy, 97.9% precision, 98.1% recall, and 98.0% F1-score, surpassing state-of-the-art baselines including CNN-LSTM, Transformer-only, and XGBoost. The robustness of the proposed system was validated under adversarial attacks, where it retained over 91.5% accuracy even against PGD perturbations, while scalability tests confirmed its ability to handle streaming loads of 80,000 messages per second with sub-200 ms latency. These results confirm that the Hybrid DeepXAI Fusion Network offers an efficient, interpretable, and secure solution for next-generation cloudbased decision-support systems.
Nancharaiah et al. (Fri,) studied this question.