Convolutional Neural Network Algorithmic Trading System

4 November 2017, 09:06
Ahmad Hassam

Convolutional Neural Networks are the latest breakthrough in deep learning.

Convolutional Neural Network have provided the breakthrough in image recognition, health and other fields.

Today there is a lot of talk on finally achieving autonomous car driving.

Can we use Convolutional Neural Networks in algorithmic trading?

I have written a blog post in which I develop a Convolutional Neural Network Algorithmic Trading System.

I have developed this model using python modules keras, tensorflow and numpy.

In this blog post, I use USDJPY 1 hour price data to predict USDJPY price 24 hours later.

The prediction was dead accurate next day.

My laptop took only 10-20 seconds to train the model with tensorflow at the backend.

I have bought a gaming laptop and yet to install tensorflow gpu.

When I do that I believe the model will take less than 1 second to make the predictions.

What this means is that we can use this model on high frequency data like 1 second and even 30 second.

Deep learning is being used extensively by the hedge funds and the big banks in building their algorithmic trading systems.

Share it with friends: