How can i unnormalize the forecasted system load outputs in Neural Networks in Matlab

I normalised and unnormalised training and test data as mentioned below and hwo can i unnormalise the forecased output to the scale of test data ?

% normalising training and test data
[pn,ps] = mapminmax(input_train);
[tn,ts] = mapminmax(target_train);
[pn1,ps1] = mapminmax(input_test);
[tn1,ts1] = mapminmax(target_test);

forecastedoutput=net(pn1);

an = sim(net,pn);
a = mapminmax('reverse',an,ts);

Answer 1

[Srivardhan Gadila]

It is recommended to normalize the entire dataset first and then split it for training and testing so that the normalization would be consistent.

Or use the same normalization settings which are used for training data to normalize the testing data:

% normalising training data
[pn,ps] = mapminmax(input_train);
[tn,ts] = mapminmax(target_train);

% normalize test data with settings used for normalizing the training data
pn1 = mapminmax('apply',input_test,ps);
tn1 = mapminmax('apply',target_test,ts);

an = sim(net,pn1);
a = mapminmax('reverse',an,ts);

Refer to the documentation of Normalize Inputs and Targets Using mapminmax & sim