Predict Output using Neural Network

Hello everyone, I have a data set which has 60 rows and 105 columns, 100 of these columns are input and 5 are outputs for the 60 elements of my data set. I would like to have a neural network which, when given the 100 input parameters, can generate the 5 output parameters based on the information that I have on the 60 elements of my current data set.

Using the neural networks pattern recognition toolbox I am able to create the neural network, but I do not know how to use it to predict other values based on just input.

In other words, how can I use neural networks to predict output based on input.

Thank you.

[Mahmood Soltani]

Daniel, I have the same question, however I am not that good with Matlab, I did not get the answers. Can you give me some help to how give the input and ask for the output out of them? Giving the exact code you entered will be great. Thanks,

[Greg Heath]

Did you notice that Daniel's post is 5 years old?

For classification examples use the commands

 help patternnet
 doc patternnet

Then search BOTH the NEWSGROUP and ANSWERS for patternnet posts

 patternnet tutorial
 patternnet greg

More examples of classification data can be obtained using

 help nndatasets
 doc nndatasets

Hope this helps.

Greg

Answer 1

[altaf adil]

You can use the function "sim" that simulates the neural network.

Just pass the trained network and test samples to the function.

Test samples can be different observations other that you have used to train the network.

Answer 2

[khu]

Nerual network toolbox creates a network based on your training dataset.

Assuming the network is named as "net" and input set for which you need output is "x", you can get the output with the following command: >> net(x)

Answer 3

[Ganesh]

You can use 75% data for testing and 25% for training the network. Select proper algorithm for training the network according to your application. You can use following code to observe "R" value with plots.

an_train = sim(net,ptr);

a_train = poststd(an_train,meant,stdt);

[X_tr,Y_tr] = size(an_train);

for i = 1:X_tr

    figure(i+X_tr)

    [m(i),b(i),r(i)] = postreg(an_train(i,:),ttr(i,:))

    k = ['Regression' int2str(i) '.dat'];

    saveas(gcf,k, 'tiffn')

end

Run the program at least 4 to 5 times and take the output with maximum "R" value.