Penny-Stock-Predictor

Tasks:

• Refine the target and make sure there are no outliers - COMPLETE
• Begin running top_board / breakout_boards updater
• Add top boards within the compiler - not till a later date...
• Create log for AWS instance to update

Questions: Q: How do I treat weekends and market holidays? A: Remove them. The message boards are much less active on those days.

GOAL:

Predict a massive change in stock price for a given small cap stock

DATA SETS :

• Web-scraped data from http://investorshub.advfn.com/
• Past stock prices and volume

MY APPROACH:

• Web scrape data from iHub
• Compile with historical stock data
• Create algorithm to determine success for target

MODEL

• Feature Space:
  • 52-week prior stock volume
  • 52-week prior iHub message frequency
  • Others (promotional emails, iHub’s “top boards”)
• Target:
  • ‘buy’ (1) : a significant stock price change is upcoming
  • ‘no buy’ (0): no significant stock price change is upcoming

Data Folder Structure

-- data

    -- raw_data
        -- ihub
            -- message_boards
            -- top_boards
            -- breakout_boards
        -- stock
            -- {raw stock data}

    -- data
        -- compiled data

SRC Folder Structure

-- src

    -- data_management
        -- ihub_data.py
            INPUT: Ticker Symbol
            OUTPUT: create/update the data/raw_data/ihub folder
        -- stock_data.py
            INPUT: Ticker Symbol
            OUTPUT: create/update the data/raw_data/stock folder
        -- compile_data.py
            INPUT: None
            OUTPUT: combined/manipulated data

    -- model
        -- model.py
            INPUT: data from the model_data folder
            OUTPUT: model

    -- data_visualization
        -- tbd

    -- web_app
        -- tbd