How To Get The Accuracy Of Independent Test By Svm?

Question

[sunflow99]

I used LibSVM to construct a prediction model with setting cross-validation to 5-fold, however, in order to avoid overfitting, an independent test is necessary.

The usage of the script "svm-train" is described as follows: svm-train -v 5 trainingsetfile

With setting the parameter -v (cross-validation), I can get the accuracy of this model. But I have to compare the performance(accuracy) of the independent test and cross-validation, so I can't set this parameter(cross-validation),however, I don't know which parameter I have to set for my independent test.

[Michael]

What do you mean by independent test? Do you mean you want to test the classifier on an additional test data set?

Answer 1

[Pappu]

You could calculate Matthews correlation co-efficient.