machine learning - Using LIBSVM grid.py for unbalanced data? -

I'm having a three square problem with unbalanced data (90%, 5%, 5%). Now I want to train a classifier using LIBSVM.

The problem is that LIBSVM optimizes the parameter gamma and cost for optimum accuracy, which means that 100% of the examples are classified as class 1, which is definitely

I have tried to modify the weight parameters, without any success

So what do I want to modify the grid in the way that it optimizes the cost and misses separation from classes rather than gamma and total accuracy for accuracy. Is there any way to do that? Or are there other scripts that can do something like this?

what the -w parameter requires unbalanced data

class 1: 5 If you have your classes:

• Class 0: 90%
• Class 2: 5% < P> You should give the following parameters for svm:

    -w0 5 -w1 90 -w2 90