Realistic Synthetic Signed Network Generation and Analysis

Designing and analyzing network science algorithms, such as node classification, link prediction, and pattern identification (communities, triangles, dense sub-graphs, cliques), require diverse real-world datasets for performance evaluation. However, these datasets are often limited and small due to privacy concerns and platform access policies. This scarcity is even more pronounced for signed networks, as negative relationship data is rarely shared publicly. This PhD thesis aims to address this problem by generating realistic synthetic signed networks using the SNSRM and SISSRM models. Preserving the mesoscopic spectral and structural characteristics of the input signed network is crucial in this process. Additionally, this thesis tackles the challenge of efficiently analyzing elementary network property of triad enumeration by developing an triangle counting algorithm capable of enumerating balanced and unbalanced triads.