A clustering algorithm is referred to as “accurate” if the majority of the data set is correctly assigned to known class labels. In such situations, the mean values of the data points in each cluster must necessarily correspond with the means of the data points in each partition. We can then compare the mean values of each cluster and each partition in order to create label transformations that can be used as a diagnostic tool in evaluating the efficacy of a clustering algorithm on a data set.
Derek Damron is currently an M.S. Statistics student at George Washington University and obtained his B.S. in Applied Mathematics from the University of Central Arkansas. Derek is currently working on a project related to cluster label assignments that might culminate into a thesis and/or conference paper. In his spare time, Derek enjoys short walks on the Navy Pier, playing video games, and people watching.
Location: Statistics Seminar Room (Rome Hall 5th Floor)
Date: February 2013