Search code examples
textsvmdocument-classification

SVM Multiclass text classification

I want to classfy News data set and training data are classified with IPTC subject code(Hierarchical classification). In my project I should use svm .

I have done all of feature extraction ,stemming,removing stop word ...

I almost have the file format required for svm multiclass :

which is like:

category  feature:value feature:value feature:value

I don't know how I should use svmmulticlass

I have also hearde about one -against-all ,I dont know what it is

please tell me whrere i can find some tutorial about it ,or how to use it

Regards

Solution

  • one -against-all, is a technique to train SVM's its in multi-label classification , for example u have "n" class label : so u create an "n" SVM and train each one on one class as +Ve label and all others as -Ve labels .

    this is what one -against-all mean . u can visit this and use there Multi-Class SVM

    http://svmlight.joachims.org/svm_multiclass.html