Archive | 2019
Exploiting Label Interdependencies in Multi-label Classification
Abstract
Multi-label classification problems very often concern multidimensional datasets, thus, performance of the method is problematic in many cases. Exploiting label dependencies may ameliorate classification results. In the paper, new effective problem transformation method which uses label interdependencies is introduced. Experiments conducted on several benchmarking datasets showed the good performance of the presented technique, regarding six evaluation metrics, including the most restricting Classification Accuracy and confirmed by statistical inference. The obtained results are compared with those obtained by the most popular problem transformation methods.