Filters, Wrappers and a Boosting-Based Hybrid for Feature Selection

Abstract

In this paper, we examine the advantages and disadvantages of filter and wrapper methods for feature selection and propose a new hybrid algorithm that uses boosting and incorporates some of the features of wrapper methods into a fast filter method for feature selection. Empirical results are reported on six real-world datasets from the UCI repository, showing that our hybrid algorithm is competitive with wrapper methods while being much faster, and scales well to datasets with thousands of features.