Unified Framework of Dimensionality Reduction and Text Categorisation

Text classification (categorization) is a supervised learning task that assigns text documents to pre-defined classes of documents. It is used to organize and manage the collection of text documents available in digital form. To accomplish the task, support vector machine (SVM) is regarded as the suitable classifier for any kind of applications. Though SVM’s computational complexity is independent of number of dimensions, still high dimensionality poses the problem of ‘curse of dimensionality’ that can be solved effectively by the process of Dimension Reduction (DR). This work contemplates on developing a framework for dimensionality reduction and text classification. A comparative analysis of the classification accuracies using two approaches viz., text classification with dimensionality reduction and text classification without dimensionality reduction completes the scope of the paper. It also evaluates the efficiency of various dimensionality reduction techniques to include one of the most coherent methods in the framework.