An Evaluation of Alternative Feature Selection Strategies and Ensemble Techniques for Classifying Music

Automatic annotation of music files is a key problem in multimedia information retrieval. In this paper we present a solution to this problem that addresses the issues of feature extraction, feature selection and design of classif ier. We outline a process for feature extraction based on the discrete wavelet packet transform and we evaluate a variety of wrapper-based feature subset selection strategies that use feature ranking based on information gain, gain ratio and principle components analysis. We evaluate four alternative classifiers; simple k-nearest neighbour and one-against-all, round-robin and featuresubspace based ensembles of nearest neighbour classifiers. The best classification accuracy is achieved by the feature subspace-based ensemble with the round-robin ensemble also showing considerable promise.