Stock Market Prediction Using Machine Learning

As part of the ML SIG Summer Project.

Project

Get Data

The Data is obtained from Quandl (restricted to the WIKI table) which requires an API key. The file get_data.py contains the necessary functions.

Usage:

python get_data.py [symbols]

For a list of available symbols for download, see: WIKI-datasets-codes.csv

Features Used

1. High-Low: It is the difference between High and Low prices of a stock for a particular day.
2. PCT_change: It calculates the percent change shift on 5 days.
3. MDAV5: It is the Rolling Mean Window calculation for 5 days.
4. EMA5: Exponential Moving Average for 5 days.
5. MACD/MACD_SignalLine: Moving Average Convergence/Divergence Oscillator. Difference between EMA26 - EMA12.
6. Return Out: Shifts the Adj. Close for stock prices by 1 day.

Models Used

1. SVM (SVC)
   • Linear Kernel
   • Polynomial Kernel
   • Radial Basis Functional Kernel
   • Sigmoid Kernel For reading: refer this
2. Ensemblers
   • Random Forest Classifier For reading: refer this

Repository

Structure of the repository

The repository houses:

• 'datasets' folder that is populated with stock data the first time script is run. To repopulate data:

  python get_data.py [quandl_symbol]

• 'research-papers' folder - the papers referred during the development of the model.
• 'environment.yml', 'requirements.txt' - See this
• 'WIKI-datasets-codes.csv' - A list of symbols to download data from Quandl.

Running for the first time?

The files environment.yml, requirements.txt make it easy to replicate the environment required for running the model.

Setting up the Environment

1. For anaconda:
   
   1. To install anaconda, refer this
      
   2. The base directory contains 'environment.yml' file. To replicate the same environment:

      conda env create -f environment.yml

2. For python3 virtual environment:
   
   1. To install virtualenv, refer this

      pip install virtualenv
      virtualenv --python=python3 ml-stock-prediction

   2. The base directory contains 'requirements.txt' file. To install the required packages:

      pip install -r requirements.txt

Getting Data

Though the datasets folder has some symbol stock prices. You can populate with more.

python get_data.py [symbols]

Running the models

You can run the model on a list of symbols supplied as command line arguments.

python main.py [symbols]