Optimizing features by correlating for concept labeling in text classification

Abstract

Unstructured form of text documents has seen a huge growth. Feature selection methods are important for the preprocessing of such text documents for dynamic text classification. Appropriate and useful features are focused during feature selection. This can decrease the cost involved while huge amount of data is dispensed out and will also amplify the next textual classifying work. This paper devised a novel geometric optimization method labeling for textual classification. An experimental study on the said geometric feature optimization method is conducted using divergent sizes of text data sets. Experimentally it is shown that how effective this method and how it is better than the tradition methods.