Barbara Ball and Clare Rodgers Jordan, College of Charleston

Using Markov Chains to Cluster Data

Abstract: Clustering data has many useful applications. In particular, search engines often employ clustering techniques to organize information on the web.  Currently, the most commonly used technique is the Laplacian method, which requires data to be in square, symmettric matrix form. However, most data is rarely in such a form, meaning information is either added or deleted when using the Laplacian method. We are attempting to improve the current selections of clustering algorithms by using Markov Chains to cluster symmetric or asymmetric matrices.

Advisor: Amy Langville (CofC)