libsvm.net is an easy way to use Support Vector Machines in your .NET projects.
Current version : 2.1.5
##How to use it First of all, if you are not familiar with SVM, I highly recommend to read this guide.
You can use libsvm.net right now, using the following NuGet package :
PM> Install-Package libsvm.net
Note : For now, C-SVC and epsilon-SVR are supported. nu-SVC and nu-SVR will be available in future versions.
The code below describe the main methods :
var prob = ProblemHelper.ReadAndScaleProblem(TRAINING_FILE);
var test = ProblemHelper.ReadAndScaleProblem(TEST_FILE);
var svm = new C_SVC(prob, KernelHelper.RadialBasisFunctionKernel(gamma), C);
var accuracy = svm.GetCrossValidationAccuracy(nr_fold);// with nr_fold > 1
for (int i = 0 ; i < test.l ; i++)
{
var x = test.x[i];
var y = test.y[i];
var predictedY = svm.Predict(x); // returns the predicted value for 'x' attributes
var probabilities = svm.PredictProbabilities(x); // returns the probabilities for each class
// Note : in about accuracy% of cases, 'predictedY' should be equal to 'y'
}Of course you can choose other kernels if you want (see KernelHelper class for more informations).
Regression methods are a little different but the spirit remains the same :
var svm = new Epsilon_SVR(prob, KernelHelper.RadialBasisFunctionKernel(gamma), C, epsilon);
double mse = svm.GetMeanSquaredError();
//...
var prediction = svm.Predict(x);Since version 2.1.0 (thanks to ccerhan), you can also save your models into xml files. This avoids having to train your model again and again when the dataset did'nt changed.
libsvm.net is provided under MIT License.
Feel free to fork and improve/enhance libsvm.net in any way your want.
If you feel that the community will benefit from your changes, please send a pull request!
libsvm.net uses the official libsvm Java library, version 3.17 translated to .NET using IKVM. It also provides helpers and classes to facilitate its use in real projects.
For more informations, see also :