Support vector machine classifier

I have input and output matrix for masonary wall panel crack. I want to classify wether the wall i cracked or not.

my input matrix is 7*576 and output is 1*576. Output matrix just have 0 and 1 value.

I need support vector machine classifier to classify my data. I cant find one class classifier.

I already tried in Back Propogation NN but the answer is not satisfactory.

[Stephan]

we now know what you want to do. We do not know what your question is. You forgot to ask a question.

[Atasham ul haq Muhammad]

I have 16 wall panels. I know the cracking patterns of all the wall panels.

i want to use AI techniques to train on some of the wall panels and predict on some wall panels.

so i have already done with NN.

Now I want to do one class classification using support vector machine.

My NN prediction code is

Input = load('NN 1 to 14 panel input.txt');

Target = load('NN 1 to 14 panel Target .txt');

Prediction_1 = load('Input_15.txt');

Prediction_2 = load('Input_16.txt');

net = newff(minmax(Input),[15,15,15,15,1],{'tansig','logsig','logsig','logsig','logsig'},'trainlm');

net.trainParam.epochs = 1000;

net.trainParam.goal = 0.0001;

net = train(net,Input,Target);

WeightBias1 = formwb(net,net.b,net.IW,net.LW);

P1 = sim(net,Prediction_1);

P2 = sim(net,Prediction_2);

I have trained the neural network on 14 panels and predicted the pattern for 2 panels.

Now I want to do the same using Support vector machine

my input matrix is 7*576 and output is 1*576

my output has only 0 and 1 value. (0 for uncracked and 1 for cracked)

So if you help me how to do this with SVM

Answer 1

[Stephan]

Hi,

i recommend to use the classification learner app to do this. The advatage is, that you dont have to write code to do this and the usage of the app seems to be clear to me. Follow the link below to learn more and for examples how to do this exactly. The whole workflow for training is described very precisly there.

• https://de.mathworks.com/help/stats/classification-learner-app.html

an additionally interesting way to learn more are the webinars - in your case check:

• https://de.mathworks.com/videos/machine-learning-with-matlab-81984.html
• https://de.mathworks.com/videos/machine-learning-with-matlab-100694.html
• https://de.mathworks.com/videos/machine-learning-with-matlab-getting-started-with-classification-81766.html

Once you have a good result on your trained models you can choose the best one and export it to matlab workspace or generate code for it, which can be modified and be used for further applications.

Best regards

Stephan

[madhan ravi]

One of the best things about Matlab is the documentations which describes everything very clearly.