what's the baseline?

[WeiFoo]

• [x] data:
  • how to pick# of classes
  • what's the percentage of each class
• [x] learner: is linearSVM is a good baseline ?

[WeiFoo]

Linear SVM, non linear SVM for text mining ??

linear SVM is as good as non linear SVM

2002 Improving Multiclass Text Classification with the Support Vector Machine

2003

[WeiFoo]

show to pick # of classes

2007 Fuzzy support vector machine for multi-class text categorization this paper pick top 4 as the class labels, and only four. see table 3

in this data set, the distribution of 4 classes are 51.3%, 33.6%, 7.5%, 7.5%

2011 A comparative study of TF*IDF, LSI and multi-words for text classification two data sets, both use 4 classes, roughly 25% for each class .