Publication details

Text document classification

Research Report

Humpolíček Jiří

publisher: ÚTIA AV ČR, (Praha 2006)

edition: Research Report 2175

research: CEZ:AV0Z10750506

keywords: text document, classification

abstract (eng):

In this report, we propose four feature selection algorithms based on the Best Individual Feature method and one based on the sequential method. After that the best method is selected for following classifier methods comparison. In this step we compare classification performance and computation expense of two classifiers based on Naive Bayes and third classifier is SVM. Classification performance is tested on the Reuters data set and Newsgroup data set. Finally we shows results on the multi-labelled subset of the Reuters data set.

Cosati: 05B, 05L