machine-edge

A modern approach to trading visualization and prediction

Redis

Redis is used as in-memory data store. You can either interact with Reids from command-line using 'redis-cli' or GUI Redis Desktop Manager. Reids supports persistence.

• The database saves to disk periodically: /var/lib/redis/dump.rdb
• The config file for redis: /etc/redis/redis.conf

Real-time market data

I have a subscription to Bloomberg where I get historical time-series and real-time market data. However, there is only Windows API. Currently, I extract these data and save them to Redis on Windows machine. Then transfer the 'dump.rdb' to Linux.

There is a Linux API, but it requires corporate subscription. We will talk about this after the prototype stage.

HP laptop

I have set up Ubuntu 16.04 LTS. I have installed the latest libraries:

Python

Please use 'python3' and 'pip3'. Ubuntu has dependencies on Python 2.x, but for this project, we will only use the latest versions.

ipython

This is a much better shell for Python. Just type 'ipython' from terminal.

jupyter notebook

Even better than ipython. Type 'jupyter notebook' from terminal.

Visualization

For the prototype, let's use Bokeh. We can bypass JavaScript completely. We will revisit and think d3.js for future development.

Some inspiration:

• http://www.thomson.co.uk/blog/wp-content/uploads/infographic/interactive-music-map/index.html
• http://evolutionofweb.appspot.com/?hl=en#/evolution/day
• https://github.com/d3/d3/wiki/Gallery

Prediction/Machine-learning

For the prototype, we can use scikit-learn

For cleaning data / data structure

Let's use pandas