Robust Detection of Regime Change

Demonstration that noise filtered correlation matrices can be used for early detection of a regime change in temporal behavior of securities. This demonstration was carried out for a portfolio of 40 S&P500 securities with just two, randomly chosen, securities undergoing a deliberately arranged regime change. Demonstration that the regime changes in the behavior of the selected two securities is manifested in very noticeable change of correlation pattern between these securities and all others, leaving correlations patterns of other securities between themselves unchanged.Demonstration that unfiltered correlation matrices don't allow detection of regime change on the same time scale