A Method Based on Support Vector Machine for Feature Selection of Latent Semantic Features
Latent Semantic Indexing(LSI) is an effective feature extraction method which can capture the underlying latent semantic structure between words in documents. However, it is probably not the most appropriate for text categorization to use the method to select feature subspace, since the method orders extracted features according to their variance,not the classification power. We proposed a method based on support vector machine to extract features and select a Latent Semantic Indexing that be suited for classification. Experimental results indicate that the method improves classification performance with more compact representation.
Qi Luo and Yuanzhi Wang
M. S. Li "A Method Based on Support Vector Machine for Feature Selection of Latent Semantic Features", Advanced Materials Research, Vols. 181-182, pp. 830-835, 2011