Analysis of Dimensionality Reduction Techniques for Effective Text Classification

Abstract

In this paper, the different dimensionality reduction techniques are presented for effective text classification. The feature selection technique and feature transformation fallowed by feature selection techniques are recommended to reduce the dimension of the features. For more compactness of data, the symbolic interval data type is recommended. The techniques are evaluated using SVM classifier and Symbolic classifier on standard benchmark datasets viz., Reuters-21578 and TDT2. The effectiveness of the techniques is verified with the performance F-1 score measure. The dimensionality reduction techniques which perform better are recommended when compared to the others.