Performance evaluation of TreeQ and LVQ classifiers for music information retrieval

Abstract

Classification algorithms are gaining more and more importance in many fields such as Artificial Intelligence, Information Retrieval, Data Mining and Machine Vision. Many classification algorithms have emerged, belonging to different families, among which the tree-based and the clustering-based ones. Such extensive availability of classifiers makes the selection of the optimal one per case a rather complex task. In this paper, we aim to address this issue by conducting extensive experiments in a music information retrieval application, specifically with respect to music genre queries, in order to compare the performance of two state-of-the-art classifiers belonging to the formerly mentioned two classes of classification algorithms, namely, TreeQ and LVQ, respectively, using a variety of music features for such a task. The deployed performance metrics are extensive: accuracy, precision, recall, Fmeasure, confidence. Conclusions on the best performance of either classifier to support music genre queries are finally drawn. © 2007 International Federation for Information Processing.