Sparse data for document clustering

Document clustering which is a part of text mining framework is used to process models and real data collection of cancer documents into several groups. A vector space model of the documents based on their key phrases are formed and called sparse matrix which contains many zero values. A sparse dimensional reduction and several methods of clustering include K-means, Self Organizing and Non-negative Matrices Factorization (NMF) are applied to the data, then the results are compared. Sparse method in dimensional reduction step using Arnoldi Method provides a better result of clustering validity twice more than standard dimensional reduction result.