Automatic Labeling of Text Document Clusters using Singular Value Decomposition

Analysis of text documents is difficult due to unstructured information it contains. Clustering of these documents helps to improve analysis under consideration. Most widely used text mining methods such as partitional algorithm k-means and hierarchical clustering methods based on linkage criterion such as single link, average link and complete link are used in this paper. The clusters are then labeled by using singular value decomposition method in a mathematical way. The labeling of the clusters makes the analyst job easier by quick capture of the cluster summary on the screen. Relative validity index is used to determine the efficiency of clustering process. It is used for estimation of number of clusters at which the process is efficient. Cluster analysis is very useful for forensic domain wherein crime investigations are performed to analyze the information from seized digital devices.