ThirstyScholar / trading-bitcoin-with-reinforcement-learning

An (unofficial) implementation of the post "Trading Bitcoin with Reinforcement Learning".
https://launchpad.ai/blog/trading-bitcoin
MIT License
97 stars 25 forks source link

Can you give your data for training? #4

Open pringwong opened 5 years ago

pringwong commented 5 years ago

When I download the data from Kaggle, I find the talib raise error for all nans inputs. And the cum return result of RL is always zero.

ThirstyScholar commented 5 years ago

Just download the data and drop the nan values.

pringwong commented 5 years ago

I have ffill nan values rather than drop nan. The cum log return rate is always zero....

LONGMIAOMIAO commented 5 years ago

I also met this condition, the RL is Zero ! My Dependecies are right below, im not sure is that matter : Numpy (1.13.3) / 1.16.0
Pandas (0.19.2) / 0.23.4 Matplotlib (2.0.0) (for plotting only) / 3.0.2 TALib (0.4.0) (to compute technical indicators described in the post) / TA_Lib-0.4.17 PyTorch (0.3.1) / 1.0.0 / python 3.6