Predicting Financial Time Series Data with Machine Learning

This is an example that predicts future prices from past price movements. Here we implement it with EUR/USD rate as an example, and you can also predict stock prices by changing symbol.

Backtest example for EUR/USD

Using daily close prices from 2008 to 2016, first 95% for training and last 5% for testing. Green and red vertical lines represent winning trade and losing trade respectively.

Installation

To run this demo, you need following environment and libraries.

• Python 2.7 (not tested on 3.x)
• Jupyter Notebook
• Scikit-learn
• numpy
• pandas
• matplotlib
• seaborn

Note: You may need extra libraries to install above.

License

MIT License, Copyright (c) 2017