Trading on Information Shocks: Network Modeling of Market Correlation Structures

Abstract

The increasingly connected nature of stock markets demands a systematic approach to the analysis of correlation structures between stocks. Using a network formulation from the correlation structures and the topology of said networks, we extract information from the network, such as topologically important nodes and hierarchical structures within subgroups of stocks. Using various selection algorithms derived from this information, we pick topologically important stocks out of this network that have had recent news releases, and then use natural language processing to predict the returns of these stocks and their nearest neighbors in order to generate a pro table trading strategy.