This repo is an exploration of using gramian angular fields to transform 1D price data to 2D image space and then performing classification on the higher-dimensional data.
I will update this readme with relevant papers / articles soon.
Note: I am exploring BTC orderflow data in this repo. The exact datasets I am using are not publically available. High quality orderflow and orderbook data for crypto exchanges is expensive to acquire, but you can find some decent sources around if you don't need something perfect.
A peer and I currently maintain a private streaming pipeline in AWS to capture and store real-time orderflow across a couple of exchanges and markets. If you are interested in this data, feel free to PM me.