Granular methods in automatic music genre classification: a case study

摘要: Classification of music files by using the characteristics songs based on its genre is a very popular application machine learning. The focus this work automatic classification granular computing methods (fuzzy rough, rough and near sets). We have proposed modified form supervised learning algorithm tolerance sets (TCL 2.0) with goal exploring scalability to well researched database composed several genres. In set method, classes are directly induced from dataset level e distance function. compared tolerance-based family nearest neighbour (NN) algorithms fuzzy (FRNN) available in WEKA platform. terms performance, accuracy TCL 2.0 identical Bayesian Networks (BN) Algorithm, comparable Sequential Minimal Optimization (SMO) Algorithm. However, average FRNN classical better than 2.0, BN SMO algorithms. For dataset, any over 90% considered good which achieved all tested classifiers work.