Characterizing and Classifying Music Subgenres

We consider the problem of subgenre classification in music datasets. We propose an adaptation of association analysis, a technique to explore the inherent relationships among data objects in a problem domain, to capture subgenres characteristics through acoustical features. We further propose to use those characteristics to engage in a pairwise comparison among subgenres when classifying a new music piece. The initial investigation on our approach is examined through empirical experiments on a number of music datasets. The results are presented and discussed, with various related issues addressed.