Data Formatting for ES-RNN

[puzzlecollector]

@kdgutier Suppose I want to train the ES-RNN on stock dataset. But because the stock market closes, I do not have data for the weekends. Would it be okay even if I have gaps in the dates for weekends and possibly some public holidays?

[kdgutier]

Hi Luke,

The ES-RNN input data needs to be clean and without missing values.

• Regarding weekends I would skip them if they are recurrent missing values like in stock data.
• Regarding public holidays I would make a simple interpolation to not mess with the ES-RNN seasonalities.

Remember that the exponential smoothing tries to learn repeated seasonal patterns, if you drift from the seasonalities the model may struggle. Also these seasonalities are treated by the model as "constant".

Because of the "constant" seasonalities and the computational cost of RNNs on very long series I recommend you to train on medium sized history windows.