Modeling and Correlation Analysis between Complex Networks

Link prediction in complex network often focuses on the single network in many fields and develops a lot of methods. However, the relationship between two dependent networks with specific global structure has hardly caught attention, but only the local topology analysis. In this paper, we aim to infer the correlation of two high dimensional dependent networks based on the similar nodes' attributes but in distinct measurement. By means of the singular value decomposition (SVD) method, we extract the model factor characteristics behind the network as latent variables to efficiently express low dimensional information, and construct linear model to estimate the parameter. We also predict and impute missing values of two networks mutually. The simulation results show the performance of model and inference quite good in the case of small sample size.