How to use trained narx network

Hi,

I am using weather parameters in order to predict solar radiation. I have used nnstart command in order to create the neural network. I first used the fitting app feature and, after saving the network, I started to use the command "y = net(inputs_data)" in order to get the predictions for a new set of inputs.It was working great. However, I realized that using a NARX network (within time series app) would give me a more accurate prediction. My issue is that I do not know how to get predictions using a new set of data once I have trained and save the NARX network as the previous command "y = net(input_data)" doesn't work in this case. I would like to precise that I only know the output values (solar radiation) for past dataset (used for the training/validation/test part) and not for the prediction process. Hence I would like to know if there is any similar command (y = net(input_data)) for NARX networks.

Thank you.

Answer 1

[Meeshawn Marathe]

Go through this documentation link. It guides you on how to train and then test the NARX model developed.

https://in.mathworks.com/help/nnet/ug/design-time-series-narx-feedback-neural-networks.html

Answer 2

[Taylor Ballard]

I'm not an expert, but I used the following line of code in its own script file after saving my network as "net" and my new input data as "IData". YPred is my new prediction.

Ypred=sim(net,IData);

Hope that helps!