PARTITIONING AND BIG GRAPH ANALYSIS

Abstract

The rapid growth of large graphs has an important role in analysing and managing their structures. The capacity to effctively analyse these graphs is crucial for many applications, including social network analysis, communication networks, and sensor networks. This thesis focuses on developing a graph partitioning based on Critical Nodes Problem (CNP) to solve several issues, including detecting communities in social networks, accurately locating the source of information in social networks, and evaluating the impact of CNP in Wireless Sensor Networks (WSNs). The fist part of this thesis describes the development of a scalable graph partitioning method that uses critical nodes to optimize vertex removal, therefore lowering network complexity while keeping critical structural characteristics and connectivity. The second part seeks to tackle the problem of community detection in social networks. We offr a unique technique that uses critical nodes to improve the quality of discovered communities. The third part emphasizes solving the problem of source detection with limited observation nodes in social networks. We propose a novel algorithm that employs critical nodes as observation nodes, it signifiantly improves source detection accuracy. The fial part of this thesis explores the impact of critical nodes in WSNs, demonstrating improved effciency, optimized performance, reduced energy consumption, and enhanced data transmission reliability. We validate the proposed approaches, according to experimental results on real-world and synthetic datasets. The results show that the algorithms outperform existing methods in terms of various performance measures, including modularity and Normal Mutual Information (NMI), accuracy, and energy consumption for WSNs.