Community identification of financial market based on affinity propagation

Abstract

Community identification in complex financial system is an important task in the exploratory analysis of stock time series. In this paper, a recently proposed message-passing-based algorithm called affinity propagation is introduced to identify stock groups. First, the similarities computed between all pairs of stocks of portfolio by considering the synchronous time evolution of their logarithm return are mapped into spatial distances. Then, the spatial distances are used to cluster the stocks into different communities of financial network via choosing appropriate preference according to affinity propagation. The results suggest that the approach is demonstrably effective in identifying multiple stock groups without any extra knowledge of stocks, and provide a meaningful economic taxonomy. © 2013 Springer-Verlag GmbH.