A multilevel tabu search algorithm for the feature selection problem in biomedical data

Abstract

The automated analysis of patients’ biomedical data can be used to derive diagnostic and prognostic inferences about the observed patients. Many noninvasive techniques for acquiring biomedical samples generate data that are characterized by a large number of distinct attributes (i.e., features) and a small number of observed patients (i.e., samples). Using these biomedical data to derive reliable inferences, such as classifying a given patient as either cancerous or noncancerous, requires that the ratio r of the number of samples to the number of features be within the range 5 < r < 10 . To satisfy this requirement, the original set of features in the biomedical data has to be reduced to an ‘ optimal ’ subset of features that most enhances the classification of the observed patients. In this paper, we propose a new feature selection technique ( multilevel feature selection ) that seeks the ‘ optimal ’ feature subset in biomedical data using a multilevel search algorithm. This algorithm combines a hierarchical search framework with a tabu search method. The framework consists of increasingly coarse forms (i.e., search subspaces) of the original feature space that are strategically and progressively explored by the tabu search method. The result of the search at any given coarse subspace is used to initialize the search at the previous less coarse subspace. We evaluate the performance of the proposed technique in terms of the solution quality, using experiments that compare the classification inferences derived from the solution found, with those derived from other feature selection techniques such as the sequential forward selection, random feature selection and tabu search feature selection. An equivalent amount of computational resource is allocated to the evaluated techniques to provide a common basis for comparison. The empirical results show that the multilevel feature selection technique finds ‘ optimal ’ subsets that enable more accurate and stable classification than those obtained by using the other feature selection techniques.