Digital Transformation Architectures For Scalable System Integration Using Cloud-Native And Deep Learning Techniques

Abstract

Today’s businesses require highly scalable and resilient digital services that can be responsive to real-time needs in a distributed and hybrid IT environment. However, traditional integration tools do not prove effective in ensuring continuous deployment, dynamic workload management, and heterogeneous system interoperability. Hence, this research presents the Digital Transformation Architecture for Scalable System Integration (DTASSI) approach aimed at solving the aforementioned problems within the context of today’s enterprise ecosystems. First, the suggested method uses data preprocessing based on Z-score normalization for achieving standardized feature scaling in relation to diverse enterprise metrics. In addition, features are extracted through the application of the Autoencoder algorithm,

which helps avoid redundancy in high-dimensional enterprise data by extracting the key system characteristics. Moreover, deep learning (DL) algorithms are applied to enhance decision-making and system optimization. To be more specific, Orca Optimization Algorithm-tuned Adaptive Graph Neural Network (OOA-GNN) is used to predict workload and latency in a highly dynamic and distributed environment by identifying complex interdependencies between various system components. OOA is nature-inspired algorithm that simulates the cooperative hunting strategies used by model's parameters. GNNs use elemental dependencies to improve prediction models.  The effectiveness of the proposed DTASSI framework is evaluated through enterprise use-case analysis focusing on scalability, fault tolerance, and operational efficiency. The experimental results demonstrate a significant improvement in system performance, achieving Cloud Native best performance of 98% accuracy, 98% precision, and 97% recall, outperforming existing baseline approaches. Additionally, Python-based implementation is used for simulation, model training, and performance evaluation, enabling scalable experimentation and reproducibility of results.