Spectral Graph Analysis Approaches for Quantifying and Enhancing Robustness in Large-Scale Communication and IoT Networks
Annotatsiya
The rapid growth of large-scale IoT networks introduces significant challenges in maintaining robustness, connectivity, and efficient communication. Existing approaches, including genetic algorithms, edge swap mechanisms, and neural network models, often suffer from high computational cost, limited scalability, or insufficient resilience under node and link failures. This study proposes a Spectral Graph Analysis (SGA)–based framework to quantify and enhance network robustness by leveraging spectral metrics such as algebraic connectivity, spectral radius, and eigenvector centrality. The proposed method identifies critical nodes and optimizes the network topology, reducing average delay to 16.9 ms, increasing throughput to 93.6 Mbps, and improving packet delivery ratio to 98.6%, resulting in an overall robustness improvement of 26% compared to baseline configurations. The framework is implemented in Python using NetworkX and NumPy on a publicly available IoT dataset from Kaggle. Results demonstrate that SGA provides a scalable, resilient, and mathematically grounded approach, offering a significant improvement in operational performance and robustness for large-scale IoT networks.