Long Short Term Memory Recurrent Neural Network Trading System

Long Short Term Memory Recurrent Neural Network (LSTM) is different from traditional neural network.

A traditional neural network uses a neurons while LSTM neural network uses memory blocks.

These memory blocks can store information for a long time before it uses it.

This makes a memory block much smarter than a neuron.

Training of these memory blocks is done through back propagation algorithm and the gradient descent algorithm.

In this blog post, I train a Long Short Term Memory Recurrent Neural Network on GBPUSD daily data.

I use Python library Keras with Tensorflow at the back end for building the LSTM model.

Once we build the LSTM model, we use it to make predicitons.

The aim is to predict next candle high, low and close.

If we can do that with a fair level of accuracy we can use this information in our daily trading.