How can I combine multiple input-target sets in NARX Neural Network to improve general model performance?

Hi there, I have been using he Neural Network Time Series toolbox to build models where I inset my 11 different inputs, and 1 output time series to produce a model. I have 6 sets of this data (let's call them A-F), and I would ultimately like to produce a model that is trained on all 6 sets of data.

Currently I am able to train a neural network on one site (e.g. A) and test on another site (e.g. B) - but to improve the model performance, I would like to train the model on e.g. 5 sets of data, and then test on 1 for validation purposes.

Any help would be gratefully received.

[christttttttophe]

Any answers on this?

Possibly divide mode would be sampletime?