Applying Deep Unsupervised Learning to understand changing financial market dynamics
GANs, Flows, and autoregressive models are used to model financial market ITCH data. This is the highest frequency message data with millions of messages per stock per day. CsStock5.py is the most succesful attempt using a autoregressive model. CsStock3.py processes the data. Modeling changing market dynamics improves the quality of message predictions, especialy during times of quickly changing market dynamics.
Note, this repository does not contain any of the required large data files.