deepakbaliga / stock-market-prediction-using-deep-learning Goto Github PK

This is a research project at the Department of Electronic & Electrical Engineering at University College London (UCL), supervised by Dr Miguel Rodrigues, named "Stock Market Prediction Using Deep Learning Techniques".

• Output data: next-day stock price trend (1 for rise, 0 for not rise)
• Input data
  • past price series (this can only generate about 50% predicting accuracy, because it carries little statistically-significant information)
  • characteristics of Twitter data related to stock (this can boost predicting accuracy because it carries more statistically-significant information)
    • daily volume of tweet messages
    • daily average sentiment score of tweet messages
• My website about the correlation between stock market and Twitter
  
  

• To run all the code, you need to pre-install the following libraries in Python 3:
  • numpy, pandas, matplotlib, scikit-learn (basic suite, most people already have)
  • pandas-datareader (for obtaining stock prices)
  • opencv (for images)
  • python-twitter (for visiting Twitter api)
  • tensorflow r0.12 (for advanced neural networks)
  • these libraries can all be easily installed through either pip or conda
• To run any code, you just need to download the folder "src", and change directory into the folder in the command line, then run the code by type "python xx.py"