Our group conducts comprehensive research on data driven methods in market microstructure and high frequency data. Recent emergence of high frequency market data thanks to automated and algorithmic trading has necessitated new methodologies to analyze and interpret such data. Classical stochastic models often struggle to incorporate all new necessary features of such data while also maintaining mathematical tractability. For instance, traditional stochastic models of financial markets often become too complicated to contain and explain market frictions, price impacts, information effects, microstructure noises, bid and ask spreads, etc., notions which can no longer be ignored in a high frequency market. Factoring these previously ignored features inside a good model is difficult, and the most likely candidate models pose challenges in analysis due to the inherent complexity of financial markets.

As alternatives to stochastic models of the usual type, we develop data driven methods to solve practical problems that traders face in today’s financial markets. Our major tools include Reinforcement Learning, Deep Q learning, Double Q Learning, GAN, Transformer, and much more.

As an academia-industry collaboration, we utilize cutting-edge machine learning techniques to tackle industrial challenges, including smart yard and smart factories.