Hands-On Machine Learning for Algorithmic Trading

How to get time series data into shape for a RNN

We will generate sequences of 63 trading days, approximately three months, and use a single LSTM layer with 20 hidden units to predict the index value one time step ahead.

The input to every LSTM layer must have three dimensions:

• Samples: One sequence is one sample. A batch contains one or more samples.
• Time steps: One time step is one point of observation in the sample.
• Features: One feature is one observation at a time step.

Our S&P 500 sample has 2,264 observations or time steps. We will create overlapping sequences using a window of 63 observations each.

For a simpler window of size T = 5, we obtain input-output pairs as shown in the following code snippet:

We can use the create_univariate_rnn_data() ...