Feature Subset Selection Using Ant Colony Optimization for a Decision Trees Classification of Medical Data

Abstract

This article describes how a feature selection is a one of the most important assignments of the preprocessing step for medical data. It can extract a subset of features from a larger set and eliminate redundant, irrelevant or noisy features. The authors can reduce the cost of diagnosis by avoiding many tests by the selection of features which are important for the prediction of disease. Applied to the task of supervised classification, the authors build a robust learning models for disease prediction. The search for a subset of features is an NP-hard problem which can be solved by metaheuristics. In this article, the hybridization between the Ant Colony Optimization and Adaboost with Decision Trees (C4.5) to improve the classification is proposed. The experiments show the usefulness of the approach.