Portfolio optimization based on network topology

The structure and dynamics of complex network systems are of current research interest. We illustrate the dependency between the network topology and its function, considering the complex financial network as a typical example. The networks are built from the full cross-correlation matrix and the global-motion one respectively, aiming at filtering the noise interference of the dynamic networks and understanding the driving mechanism of different interactions. Dynamic structural features of the core and periphery nodes are investigated, and it is demonstrated that the peripherality in a network can be used as an indicator for identifying the optimal assets. With the network filtering approach and peripherality measure, portfolios with different performances are constructed. Compared to the full cross-correlation matrix, the global-motion one shows significant advantages in the portfolio optimization, and the underlying mechanism is carefully analyzed. These methods are also with potential significance to the understanding of other social, biological and transport systems.