search

ZhengyaoJiang / PGPortfolio

PGPortfolio: Policy Gradient Portfolio, the source code of "A Deep Reinforcement Learning Framework for the Financial Portfolio Management Problem"(https://arxiv.org/pdf/1706.10059.pdf).
GNU General Public License v3.0
1.74k stars 750 forks source link

Stock Application #35

Closed joaosalvado10 closed 6 years ago

joaosalvado10 commented 6 years ago

Hello, I am interested in applying this solution to stock markets instead of cryptocurrencies. I Have a csv file containing the Open High Low Close of each one of my stocks. Let's say I Have 5 stocks for example. I would like to know how should I proceed in order to adapt this solution to the stock market instead.

Really nice project. Thank you

ZhengyaoJiang commented 6 years ago

Hi there, I think you need first modify "globaldatamatrix.py". The most important interface is get_global_panel(). The csv should be transformed into a pandas panel whose axis are "[feature, coin, time]".

zhengyao

joaosalvado10 commented 6 years ago

@ZhengyaoJiang hello, I have been trying to modify it still I could not find a way to do it. Do you plan to do add this type of feature?

Also, I have some questions. Where is the main loop of Reinforcement learning?

And how would I use the model to predict the action it if I receive new prices?

ZhengyaoJiang commented 6 years ago

Where is the main loop of Reinforcement learning?

The main loop of training is in tradertrainer.

And how would I use the model to predict the action it if I receive new prices?

You can use the decide_by_history method of nnagent

chankeypathak commented 5 years ago

@ZhengyaoJiang , @joaosalvado10 Has this been implemented? If yes then could you please share an example of the same? Thanks.