How to test using Matlab IDE?

Hi,

I have got a training data. I have used the 'train' option to tain the SVM kernel. Now I have a test data also. How i can test SVM kernel using test data. Please guide me.

Is there a test option or not in Matlab IDE?

Please guide me how to test using the test data through Matlab IDE using SVM Kernel (Guassian).

Zulfi.

[madhan ravi]

zulfiqar khan's answer moved here for consistency

Hi,

I found a link on Matlab which says:

Export the trained model into a net/ weights. Use this on the test data and get regression

Can some body please guide me What is meant by net/weights & how to use it on test data & how to get regression?

Zulfi.MathWorks

[madhan ravi]

zulfiqar khan's answer moved here for consistency

Hi,

I found a link:

MathWorks

which says:

"Make Predictions for New Data

After you export a model to the workspace from Classification Learner, or run the code generated from the app, you get a trainedModel structure that you can use to make predictions using new data. The structure contains a classification object and a function for prediction. The structure allows you to make predictions for models that include principal component analysis (PCA).

1. To use the exported classifier to make predictions for new data, T, use the form:

yfit = C.predictFcn(T)

"

But when I am exporting the model (without generating the code) my workspace still contains the confusion matrix (i.e the positive predict values).

Please guide me how to clean the workspace and do the prediction using the new data.

Zulfi.

[madhan ravi]

Please don't use the answer section for commenting use comment section instead. Thank you for understanding.

[Zulfiqar Khan]

Hi,

Sorry for using answer section for commenting. In the above post, I confused Editor with workspace. Now I found my model uploaded in the workspace. I have clicked the link "How to predict using an exported model". It tells me to use:

yfit = GTM_ACC62p8.predictFcn(T)

where T is the data.

Should I write the above statement in the Editor? Editor is showing property & value (image attached). What does it mean?

Also I dont know how to import 'T' because at this point I can't see the import option. This is because "variable" tab is activated. Should i click on Home tab to import data & then click on the App tab and then click on run.

Some body please guide me.

Zulfi.

Answer 1

[Alok Nimrani]

Hi Zulfiqar,

If you want to train a SVM kernel and use it later for testing, please have a look at the following link which describes the use of a Gaussian kernel for training a classificaiton model and then using 'predict' function to test the kernel: Gaussian Kernel Classification Model

However, if your goal is to perform regression, the following link might be helpful which describes the use of Gaussian kernel training for a regression model and then testing this model using 'predict' function: Gaussian Kernel Regression Model

Kindly go through the examples in these links. Hope this helps.

Regards,

Alok

[Zulfiqar Khan]

Hi,

Thanks my friend for your response. Your solution is not working. I tried the following:

Label = predict(trainedModel, testdata);

Error using predict (line 84)

No valid system or dataset was specified.

I already have dataset "testdata" imported in my workspace. Why its saying its not valid?

Please guide me.

Zulfi.

[Alok Nimrani]

Hi,

Sorry for the delayed response. Please note that the predict function can be used with either a RegressionKernel model object or a ClassificationKernel model object as specified in the input arguments of the links shared in the above answer. So, the error that you are getting is probably because your model is not belonging to one of these types.

Try creating a RegressionKernel or a ClassificationKernel model object and then use the predict function. If this does not resolve the issue, please share a sample code so that I can reproduce the issue.

Regards,

Alok