Abstract
We propose the application of a high-speed maximum likelihood clustering algorithm to detect temporal financial market states, using correlation matrices estimated from intraday market microstructure features. We first determine the ex-ante intraday temporal cluster configurations to identify market states, and then study the identified temporal state features to extract state signature vectors (SSVs) which enable online state detection. The SSVs serve as low-dimensional state descriptors which can be used in learning algorithms for optimal planning in the high-frequency trading domain. We present a feasible scheme for real-time intraday state detection from streaming market data feeds. This study identifies an interesting hierarchy of system behaviour which motivates the need for timescale-specific state space reduction for participating agents.
Acknowledgements
The conclusions herein are due to the authors and the NRF accepts no liability in this regard. We also thank the Fields Institute for Research in Mathematical Sciences and the University of Toronto for hosting the first author while much of the work for this paper was completed.
Notes
No potential conflict of interest was reported by the authors.
1 This form of the price model ensures that the self-correlation of a stock is one and independent of the cluster coupling. This can be seen by computing the self correlation and using that clusters and stock unique process are unit variance zero mean processes(3)
This is not a unique choice, another possible choice often used is(4)
3 This form of the price model ensures that the self correlation of a stock is one and independent of the cluster coupling. This can be seen by computing the self correlation and using that clusters and stock unique process are unit variance zero mean processes(A2)
This is not a unique choice, another possible choice often used is(A3)
4 The trace of the correlation matrix for each cluster s can be verified from the eigenvalues(B17)