Multivariate Time Series Forecasting with LSTM?


Im trying to use Keras on a dataset with multiple features,
but get this message: “For non-scalar data values, only single column selection is supported.”

Isn’t this possible with the keras nodes, or am i doing something wrong?

Input example:

1 Like

Hi Kibec,

can you provide a screenshot of your learner/executor dialog that shows the assignment of table columns to input tensors?



Hi Nemad,


Ah yes, we currently don’t support nested lists in a straight forward way although you can use the Image Processing Plugin to have such multidimensional inputs.
However, in your case your input tensor is one-dimensional so you can simply flatten your list into a single list or multiple columns by using the Split Collection node.
Please also note that all your training examples need to have the same shape i.e. number of elements, even if the shape is theoretically dynamic (contains “?”).



1 Like

Thank you for the quick and helpfull reply!

1 Like

This topic was automatically closed 90 days after the last reply. New replies are no longer allowed.