ML-algotrade

Project Title: Algorithmic Trading with Machine Learning

1. Project Proposal

Group project for Baruch MTH 9899 - Big Data II: Machine Learning

1.1 Project Member

• Weiyi Chen, weiyi.chen@baruchmail.cuny.edu
• Yifan Cui, yifan.cui@baruchmail.cuny.edu
• Chengbin Wang, chengbin.wang@baruchmail.cuny.edu
• Haotian Wu, haotian.wu@baruchmail.cuny.edu

1.2 Project Idea

Use different machine learning algorithms to modify traditional technical strategies, and backtest with research in cycle to find a good algorithm,

• Machine Learning algorithms
  • Nearest Neighbors, http://scikit-learn.org/stable/modules/neighbors.html
  • Random Forests, http://scikit-learn.org/stable/modules/ensemble.html#forests-of-randomized-trees
  • Support Vector Machine, http://scikit-learn.org/stable/modules/svm.html
• Technical Strategies
  • Bollinger Band
  • Momentum
  • Rotated Relative Graph We use Introduction to machine learning as our guide to understand the algorithms and Evidence-based technical analysis to learn technical strategies.

1.3 Data Set and Software: Quantopian

Quantopian is our main research and backtesting platform, we research algorithmic trading ideas and explore curated financial data with the assistance of machine learning algorithm in this hosted research environment. Quantopian supports flexible data access, custom plotting, and post-hoc analysis on backtest.

For convenience to verify our machine learning algorithms by cheating ('cheat' is like using today's data to forecast today), we also support local access to data from Yahoo, with event study and backtesting tools from Python module QuantSoftware Toolkit. In concludes,

• Data input
  • Quantopian, or Yahoo
  • Equity Universe: adjusted close of S&P500 index equity component
  • Frequency: daily
• Software
  • Quantopian, a platform to build, test, and execute trading algorithms.
  • QuantSoftware Toolkit, a modified Python-based open source software framework designed to support portfolio construction and management

1.4 Programming Language

Python. The requirement of this course suggests Matlab, but we request for using Python because our preferred platform Quantopian is using Python and the Python module QSTK we will be using is also in Python.

The main module we will be using is scipy-learn (http://scikit-learn.org/stable/supervised_learning.html#supervised-learning), which contains implemented Machine Learning algorithms.

1.5 Books and Papers reference

• Alpaydin, Ethem. Introduction to machine learning. MIT press, 2014.
• Aronson, David. Evidence-based technical analysis: applying the scientific method and statistical inference to trading signals. Vol. 274. John Wiley & Sons, 2011.

2. Installation Guide

Repository installation

2.1 System Requirements

Make sure your system meets these requirements:

• Operating system: MacOS 10.9 10.10 (it has been tested successfully on these)
• RAM: 2GB.
• Disk space: 2GB

2.2 Install Command Line Tools

• Open terminal, type “xcode-select --install” in terminal (without quotes)
• A pop-up windows will appear asking you about install tools, choose install tools, wait install to finish

2.3 Install Homebrew

ruby -e "$(curl -fsSL https://raw.githubusercontent.com/Homebrew/install/master/install)"
brew tap samueljohn/python
brew tap homebrew/science

2.4 Install gcc and its libraries

brew install gcc

2.5 Install Python and its modules

brew install python

pip install virtualenv
pip install nose
pip install pyparsing
pip install python-dateutil

pip install numpy
pip install scipy
pip install matplotlib

pip install pandas
pip install scikits.statsmodels
pip install scikit-learn
pip install QSTK

2.6 Install Sublime IDE and Fix Mac Path (other systems no need to fix)

brew install sublime
git clone https://github.com/int3h/SublimeFixMacPath.git ~/Library/Application\ Support/Sublime\ Text\ 2/Packages/FixMacPath

2.7 Clone AlgoTrading and Overwrite QSTK package

git clone https://github.com/weiyialanchen/AlgoTrading.git
cp -r AlgoTrading/* /usr/local/lib/python2.7/site-packages/QSTK/
rm -r AlgoTrading/

2.8 Set up alias to working directory

Note that your working directory is /usr/local/lib/python2.7/site-packages/QSTK/, to make alias

subl ~/.bash_profile

Append the following line to the file

alias AlgoTrading='cd /usr/local/lib/python2.7/site-packages/QSTK/'

Save, exit and restart the terminal. Next time when you want to go to the working folder and open it, just run

AlgoTrading
subl ./