TensorFlow implementation of Z. Hu et al. "Listening to Chaotic Whispers: A Deep Learning Framework for News-oriented Stock Trend Prediction", WSDM 2018

• Main components

  • TensorFlow 1.4.0
  • Numpy
  • Scikit-learn

• Dataset

  • Stock prices and tweets
    • Yumo Xu and Shay B. Cohen "Stock Movement Prediction from Tweets and Historical Prices", ACL 2018.
    • Copy https://github.com/yumoxu/stocknet-dataset/tree/master/price/preprocessed/* files to {PROJECT_PATH}/data/price/preprocessed/
    • Copy https://github.com/yumoxu/stocknet-dataset/tree/master/tweet/preprocessed/* files to {PROJECT_PATH}/data/tweet/preprocessed/
    • 87 stocks (S & P 500)
    • 31 Dec 2013 ~ 31 Dec 2015

• Working directory setting

$ export PYTHONPATH=$PYTHONPATH:$(pwd)

• Word Representation
  • Download https://nlp.stanford.edu/data/glove.twitter.27B.zip
  • Extract to data/

• Word Representation

  • fastText (ref. https://github.com/facebookresearch/fastText#building-fasttext-for-python)

    • Installation

    $ git clone https://github.com/facebookresearch/fastText.git
    $ cd fastText
    $ pip3 install .

    • wiki english folder set
      • ~/common/fasttext/wiki.en.bin
      • Download - https://s3-us-west-1.amazonaws.com/fasttext-vectors/wiki.en.zip

  • BERT

    • BERT-Large, uncased, whole word masking

      • https://storage.googleapis.com/bert_models/2019_05_30/wwm_uncased_L-24_H-1024_A-16.zip

    • BERT tokenization

      • https://github.com/google-research/bert#tokenization