SISSRM: Sequentially Induced Signed Subnetwork Reconstruction Model for Generating Realistic Synthetic Signed Networks

Privacy is one of the major concerns in the availability of large-scale real-world network datasets for various applications. Therefore, we require network models that can generate realistic synthetic networks of a very large scale that are capable of preserving patterns of given structural property in its evolution period. In this article, we proffer a novel network reconstruction model for signed networks, i.e., sequentially induced signed subnetwork reconstruction model (SISSRM), that is able to preserve the distribution of degrees, obsolescence, unbalanced, and balanced triangles, correlation among diverse triad types, spectral radius, degree correlation, and network balancedness during its growth process. SISSRM reproduces different structural properties of a given real-world signed network more accurately as compared to the considered state-of-the-art network models. The extensive experimentation on nine real-world empirical networks validates the significance of our proposed model in the generation of realistic synthetic signed networks.