libsvm - 使用 SVM 进行文档分类

Question

哪个是文档分类项目的最佳 SVM 算法？

Answer 1

Document classification tasks typically feature high dimensional, sparse input spaces when you use a bag-of-words vectorization of the documents (typically in tf-idf form). This is an ideal scenario for linear SVM. An important publication which inspired many researchers to use linear SVM for document classification is "Text Categorization with Support Vector Machines" by Thorsten Joachims, available here.

It is recommended to use specialized linear SVM packages (such as LIBLINEAR, SVMPERF, PEGASOS) for this type of task since they are much much faster than general SVM packages when training a linear SVM.