Indian Journal of Science and Technology

Abstract

Correlation Preserving Indexing (CPI) can discover the intrinsic structures implanted in high-dimensional document space. To predict the result of one variable based on another variable is not suitable for all the situations since two variable prediction problems takes places. In this paper, Directed Ridge Regression is introduced to predict two or more variables which are highly correlated in high dimensional document space. Directed Ridge Regression is a statistical technique to estimate the relationship among the variables based on the Eigen values to find the similarity between the documents. The directed ridge estimator alters the diagonal elements of the Eigen values. The objective of the Directed Ridge Regression is to achieve efficient document clustering in similarity measure. Experimental results shows that compared to Correlation Preserving Indexing, the Directed Ridge Regression achieves efficient document clustering.

Keywords: Correlation Similarity Measure, Directed Ridge Regression, Document Clustering, Latent Semantic Indexing