All Times EDT
Keywords: Logistic regression model, Covariance structure, Market state, High frequency, Asset management
When modelling multivariate financial data, it is compounded that the covariance structure changes with time. Previous work includes time series models switching with market states and an alternative graphical model when changes occur at random time. Based on the implementation from these models that the addition or deletion of an edge changes with the market states shifting, we use a logistic regression model to learn the stock market states. Applications are designed from several financial intuitions. One is the liquidity amplification in extreme states. The other is low and high frequency data difference. We demonstrate a hedging strategy as a combination of active and passive asset management and find its superior performance in both cases of time frequencies.