This is the final graduation capstone project were I explain the approach to develop a Reinforcement Learning agent designed to learn to trade Bitcoin.
Find also the link to the companion repository that contains the code for the deep-q trading agent.
I have used a dataset consisting of OHLCV prices from BitMEX, quite a traditional approach. As order book historical dataset is available from sites like algorithmic.ch, it would be great to see how does the model perform trying to predict bitcoin prices from the order book perspective.