Date of Award


Document Type


Degree Name

Master of Science (MS)


Computer Science

First Advisor

Dr. Parvathi Chundi

Second Advisor

Dr. Mahadevan Subramaniam

Third Advisor

Dr. Abhishek Parakh


Multi-label classification (MLC) is one of the major classification approaches in the context of data mining where each instance in the dataset is annotated with a set of labels. The nature of multiple labels associated with one instance often demands higher computational power compared to conventional single-label classification tasks. A multi-label classification is often simplified by decomposing the task into single-label classification which ignores correlations among labels. Incorporating label correlations into classification task can be hard since correlations may be missing, or may exist among a pair or a large subset of labels. In this study, a novel MLC approach is introduced called Multi-Label Classification with Label Clusters (MLC–LC), which incorporates label correlations into a multi-label learning task using label clusters. MLC–LC uses the well-known Cover-coefficient based Clustering Methodology (C3M) to partition the set of labels into clusters and then employs either the binary relevance or the label powerset method to learn a classifier for each label cluster independently. A test instance is given to each of the classifiers and the label predictions are unioned to obtain a multi-label assignment. The C3M method is especially suited for constructing label clusters since the number of clusters appropriate for a label set as well the initial cluster seeds are automatically computed from the data set. The predictive of MLC–LC is compared with many of the matured and well known multi-label classification techniques on a wide variety of data sets. In all experimental settings, MLC–LC outperformed the other algorithms.


A Thesis Presented to the Department of Computer Science, and the Faculty of the Graduate College University of Nebraska In Partial Fulfillment of the Requirements for the Degree Masters of Science University of Nebraska at Omaha. Copyright 2018 Dilanga Lakshitha Bandara Abeyrathna, Galapita Mudiyanselage.