How to display Confusion matrix of Testing ,Training and Validation without using nprtool

Hello Everyone ,

I want to display confusion matrix of Testing , Training and validation through code .

plotConfusion(Target, Output) only displays the overall confusion matrix .

 i can only show the three matrices first time i train the network using newpr. How to display them again ?

Answer 1

[Greg Heath]

net = patternnet(x,t,H); [net tr ] = train(net,x,t);

The training record in the structure tr contains the indices for the trn/val/tst subsets.

Therefore, they can be used to obtain separate performance results.

Hope this helps.

Greg

[Saira]

Thank You Greg !

This was really helpful , but i'm still unable to get the same confusion matrix .

Here is my code after you suggestion :

one thing more i'm using Matlab 7.9 , so it doesn't have patternet function .

for i = 1:540 % 540 is total number of input samples

for j = 1:324 % 324 is number of samples used only for training ,remaining 108 for testing & 108 validat

if(i==tr.trainInd(1,j))

trainOut(:,k)=output(:,tr.trainInd(1,j)); % output = sim(net,input)

% Here we are taking corresponding output values of the training indices.

k=k+1;

end

end

end

%.Similarly i got corresponding Target values using the same code above with a little change

for i = 1:540

for j = 1:324

if(i==tr.trainInd(1,j))

trainTar(:,k)=Tar(:,tr.trainInd(1,j)); % getting corresponding target values

k=k+1;

end

end

end

% Then i plot confusion matrix for Training Data only

plotconfusion(trainTar,trainOut)

[Rina Blomberg]

% x = inputs, t = targets, y = outputs

% Train the Network

[net, tr] = train(net, x, t);

% Training Confusion Plot Variables

yTrn = net(x(:,tr.trainInd));

tTrn = t(:,tr.trainInd);

% Validation Confusion Plot Variables

yVal = net(x(:,tr.valInd));

tVal = t(:,tr.valInd);

% Test Confusion Plot Variables

yTst = net(x(:,tr.testInd));

tTst = t(:,tr.testInd);

% Overall Confusion Plot Variables

yAll = net(x);

tAll = t;

% Plot Confusion

plotconfusion(tTrn, yTrn, 'Training', tVal, yVal, 'Validation', tTst, yTst, 'Test', tAll, yAll, 'Overall')

[Supriya Pahwa]

yTst = net(x(:,tr.testInd)) when i use this command, i get an error. Subscript indices must either be real positive integers or logicals.