A network-based method for detecting critical events of correlation dynamics in financial markets

In recent years, the dynamics of the financial correlation matrix has been widely studied. This study applies the Frobenius distance-based kNN (k nearest neighbour) network to analyse the time-varying characteristics of the correlation matrix, especially during the period of drastic change. We use the influence-strength (IS) index to detect when the correlation matrix structure changes dramatically. Based on the data from the US stock market, we tested the effectiveness of the method. The IS-based method accurately detects some important events from the 2008 crisis to the 2020 crisis. Our calculations indicate that IS-based analysis provides an effective tool for analysing financial correlation dynamics.