Observation space: Three vectors (v1, v2, v3) that represent features extracted from raw data feature extractors (i.e. financial news articles, stock price history, financial fundamentals).
Action space:
Version 1: buy, sell or hold (0,1 or 2) for each stock in the portfolio.
Version 2: buy or sell (1 or 0) for each stock in portfolio.
Version 3: buy, sell and choice of asset. How to incorporate the stock-tick choice as an action by the RL agent?
Version 4: vector of length equal to number of stocks that can be traded at once. Each element has value between -inf to inf representing the value of the transaction. Negative values indicate a sell and positive values indicate a buy. Value is proportional to the number of shares in the transaction (value/asset_price = number of shares). 0 or < 1 share would indicate a "hold"
Could extend this openAI gym environment -> https://awesomeopensource.com/project/AminHP/gym-anytrading
State Space: (price, number of shares, cash).
Portfolio value = price*n_shares + cash = reward.
Observation space: Three vectors (v1, v2, v3) that represent features extracted from raw data feature extractors (i.e. financial news articles, stock price history, financial fundamentals).
Action space: