• Community Discovery
• Conductance
• Egonet
• Friendship
• GELSE
• LECD
• Seed Expansion
• Social Networks Analysis
• TECD

Network analysis has become a widely studied topic in both academic and non-academic circles. The expressive power of graphs allows for the effective representation of diverse contexts, making network analysis applicable in areas ranging from medicine to social science. One crucial activity within this field is community discovery, which involves identifying clusters that share similar characteristics. Over the years, various algorithms have been developed, initially targeting entire networks and later focusing more on local analyses due to computational challenges and data limitations.
This thesis presents three community discovery methods: one local LECD, one global GELSE, and one temporal TECD. The primary focus is on LECD, which utilizes egonets, conductance, and the concept of community cohesion, friendship and lite friendship. LECD operates in two phases: the first phase identifies core area, and the second phase expands these core nodes while minimizing conductance. GELSE revisits LECD and attempts to extend the local method to a global approach without losing the focus on local tactics. TECD extends LECD to temporal networks using an instant optimal approach.
The discussion details the metrics and measures used by the algorithms, followed by pseudocode and complexity analysis. Experiments on both local and global networks evaluate the algorithms' efficiency, stability, and performance in both real and synthetic contexts. Results indicate that LECD is a stable and effective alternative to existing methods, while GELSE represents an interesting experiment, albeit with some performance limitations. TECD, on the other hand, has proven to be simple and effective for monitoring the seed's community over time.