Trader which take into account twitter sentiment on specific topic and Technical Analysis to decide to go Long or Short.
Currently this package has three main pilars: (1) Feature engineering on tweet sentiments and on prices using Technical Analysis; (2) Reinforcement learning environment for trading and (3) Jupyter Notebook which join all this parts and train an RL DNN model.
API Documentation here
In order to install the required packages, please create an Conda environment from environment.yml
using the following command:
conda env create -n <ENV NAME> -f environment.yml
Besides, to configure the Tweepy streaming, you will need to add some tokens on settings.py.default
. There are plenty of guides on Internet on how to get these Tweeter API tokens (here you can find an example). Then add them on settings.py.default
and once done, please rename the file by settings.py
.
In order to run Jupyter Notebook presented here, some CSV files are required. These have been extracted from Kaggle and they have to be allocated on data/source/
folder. The CSV files used are:
- Tweets about Bitcoin: https://www.kaggle.com/alaix14/bitcoin-tweets-20160101-to-20190329
- Bitcoin prices: https://www.kaggle.com/mczielinski/bitcoin-historical-data#bitstampUSD_1-min_data_2012-01-01_to_2019-08-12.csv
- Other historical prices: https://www.dukascopy.com/trading-tools/widgets/quotes/historical_data_feed