Performance Analysis of a Gaussian Mixture based Feature Selection Algorithm

Abstract

Feature selection for clustering is difficult because, unlike in supervised learning, there are no class labels for the data and, thus, no obvious criteria to guide the search. The work reported in this paper includes the implementation of unsupervised feature saliency algorithm (UFSA) for ranking different features. This algorithm used the concept of feature saliency and expectation-maximization (EM) algorithm to estimate it, in the context of mixture-based clustering. In addition to feature ranking, the algorithm returns an effective model for the given dataset. The results (ranks) obtained from UFSA have been compared with the ranks obtained by Relief-F and Representation Entropy, using four clustering techniques EM, Simple K-Means, Farthest-First and Cobweb.For the experimental study, benchmark datasets from the UCI Machine Learning Repository have been used.