Output details
11 - Computer Science and Informatics
University of Surrey
Hybrid ACO and TOFA feature selection approach for text classification
<24>The novelty of this paper is the introduction of a new technique to combine two optimisation methods, ACO and TOFA, for automatic feature selection for classification. This results in more accurate classification than each method individually. The method has demonstrated its success in text classification and is also applicable to images. We are currently extending it to medical image analysis: when clinical signs vary from patient to patient, the difference between initial presence of a clinical sign and a normal case is very subtle, whilst the classification task itself requires high precision, and is improved by this method.