Document Clustering based on mutual information and PCA subspace

Mutual information is a criterion widely used in statistical language modeling word associations and feature selection. Principle component analysis (PCA) is a statistical technique for unsupervised dimension reduction. K-means clustering is commonly used data clustering for unsupervised learning tasks. In this paper, we first select features from native feature space by mutual information to reduce data dimension, then decompose the covariance matrix of new word-document matrix via Singular Value Decomposition (SVD), finally, K-means is used to cluster in PCA subspace. The clustering result is analyzed comprehensively under different condition.