Time series forecasting example is mysterious

I’m having trouble understanding the time series forecasting example in the examples. It needs more documentation, and more comments in the code.

  • Why does it define the width of the time series window as 20 and then not use it in the model? How is that information provided to the LSTM? I guess it doesn’t matter?
  • What is up with the magic numbers? It’s got “51” hard coded all over the place with zero explanation of what it is. Why does the 51 matter? How was that value selected? Why not 50?
  • It simply doesn’t explain what it’s doing. What’s the difference between LSTM and LSTMCell? Why does the model use two LSTMCells and not an LSTM? I guess that means LSTM is not suitable for time series prediction?

Has anyone figured this code out, or can explain what it’s doing? What is that 51 all about? Why doesn’t the length of the time series matter?

1 Like

It would be great if you could edit your post to provide a link to the example you mention, I can’t even find a ts forecasting example ;-).Thanks. [EDIT: found it here.]