Classification Inductive Rule Learning with Negated Features

This paper reports on an investigation to compare a number of strategies to include negated features within the process of Inductive Rule Learning (IRL). The emphasis is on generating the negation of features while rules are being "learnt"; rather than including (or deriving) the negation of all features as part of the input. Eight different strategies are considered based on the manipulation of three feature sub-spaces. Comparisons are also made with Associative Rule Learning (ARL) in the context of multi-class text classification. The results indicate that the option to include negated features within the IRL process produces more effective classifiers. Key words: Rule Learning, Negation, Multi-class Text Classification

• Type: Journal
• Year: 2010
• Where: ADMA
• Authors: Stephanie Chua, Frans Coenen, Grant Malcolm
• Keywords: ADMA 2010, Data Mining, Multi-class Text Classification, Negation, Rule Learning