Multi - Classification using SVM

kailashbhar

Hi, I am working to classify documents using SVM. I have prepared an excel file having two columns. One column is document name and another column have major key words for a  document. There are about 150 rows of data in file. I have mnaully labeled this data in 8-10 categories. Now, I want to to calculate performance of SVM using this data. All data in file is of text type. Could you please help how can I use SVM for this problem using Rapidminer. Thanks in advance.

MariusHelf

Does each document belong to exactly one category? Then you can use e.g. the operator Polynominal by Binominal Classification, which creates N model for N classes/categories, where the i-th model decides if a document belongs to the i-th category or not. In literature this method is referred to as 1-vs.-all classification.

Please note, however, that your results will probably not be as good as expected, since only 150 documents for 10 categories is usually way to few data.

Best regards,
Marius