An experimental investigation on PCA based on cosine similarity and correlation for text feature dimensionality reduction

Principal component analysis (PCA) is a commonly used method for feature extraction and dimensionality reduction. This paper proposes PCA based on similarity/correlation criteria instead of covariance to gain low-dimensional features with high performance in text classification. Experimental results have demonstrated the advantages and usefulness of the proposed method in text classification in high-dimensional feature space, in terms of the number of features required to achieve the best classification accuracy.