LSTM Composite implementation in TensorFlow described in [1] by combining an autoencoder and a Seq2Seq for its predictor
Extracted n-day market pattern using k-means from Scikit-Learn
The following are the required packages to run the models.
We recommend setting up a virtual environment with Python > 2.7.x (tested on 2.7.10):
virtualenv -p python venv
source venv/bin/activate
Install all required packages by running:
pip install -r requirements.txt
Run the notebooks with jupyter notebook
In addition you can use this checkpoint to initialize your model for the demo (more checkpoints to come).
The model is trained on load_OHLC_no_vol()
with hidden_size=[128], encoder_steps=24, decoder_steps=24
.
Additional data is required for kmeans-demo.ipynb and available upon request.
This project is licensed under the MIT License - see the LICENSE.md file for details