Portfolio Selection Based On Social Network Theory

In this paper, portfolio selection based on the social network clustering and centrality ranking method and its investment efficiency is discussed. Firstly, assuming each stock is treated as a node, one can regard a stock market as a social network. Then multiple in-dexes of the stock (technical or financial indexes) are used to calculate the stock similarity,which is also chosen as the joint weight of the nodes. The initial stock social network is established. A spectral clustering is made for the social network to classify stocks based on Louvain algorithm. Secondly, social network node degree is used to optimize security social network by the power-law distribution. And then the centrality indexes of the stock are calculated. The improved TOPSIS method based on centrality indexes of the stock is used to create portfolio. Finally, empirical tests are conducted to assess portfolio selection. The empirical results show that the method can greatly improve the efficiency of investment and reduce the investment risk.