Analysis of Sparsity in a Support Vector Machine Based Feature Selection Method

Text classification is an important and well studied area of pattern recognition, with a variety of modern applications in natural language documents; we classify text documents into a set of predefined categories. Under the sparse model documents are represented by sparse vectors, where each word in the vocabulary corresponds to one coordinate axis. In a large collections of documents, both the time and memory required for training classifiers connected with the processing of these vectors may This calls for using a feature selection method, not only to reduce the number of features but also to increase the sparsity of document vectors.