An MCMC Feature Selection Technique for Characterizing and Classifying Spatial Region Data

We focus on characterizing spatial region data when distinct classes of structural patterns are present. We propose a novel statistical approach based on a supervised framework for reducing the dimensionality of the initial feature space, selecting the most discriminative features. The method employs the statistical techniques of Bootstrapping simulation, Bayesian Inference and Markov Chain Monte Carlo (MCMC), to indicate the most informative features, according to their discriminative power across the distinct classes of data. The technique assigns to each feature a weight proportional to its significance. We evaluate the proposed technique with classification experiments, using both synthetic and real datasets of 2D and 3D spatial ROIs and established classifiers (Neural Networks). Finally, we compare our method with other dimensionality reduction techniques.