Classificacao Automatica de Generos Musicais Utilizando Metodos de Bagging e Boosting

This paper presents a study that uses meta-learning techniques to the task of automatic music genre classification. The meta-learning techniques we used are Bagging and Boosting. In both cases the component classifiers used in both approaches are Decision Trees, k-NN (k nearest neighbors) and Naive Bayes. The experiments were performed on a dataset containing 1,000 songs with 10 different genres. The achieved results show that the Bagging approach is promising while the Boosting approach seems to be inadequate to the problem.