[BibTeX] [RIS]
Examining Label Intersections in Pairwise Multilabel Classification
Type of publication: Mastersthesis
Citation: ba:gasiorowski
Type: Bachelorarbeit
Year: 2015
Month: September
School: TU Darmstadt, Knowledge Engineering Group
URL: http://www.ke.tu-darmstadt.de/lehre/arbeiten/bachelor/2015/Gasiorowski_Tomasz.pdf
Abstract: Due to the overwhelming amount of data being processed nowadays, the importance of auto- mated data classification is rising. Classifiers are relevant for many industries as they handle topics such as medical image analysis, internet search queries and email spam filtering. Typical tasks can be solved through multiclass classification, which involves assigning one of multiple classes to an instance. A variant of the traditional multiclass classification problem is multil- abel classification. In this setting, classes are not mutually exclusive and samples can belong to multiple classes simultaneously. The dependencies between classes, particularly their over- lapping areas, is the reason why this is a challenging problem. Classification through pairwise decomposition is one of the leading methods for solving multilabel classification. In this method we transform the multilabel problem into single-label problems through learning classifiers for each pair of labels and combining their outputs to receive the end result. In this thesis we will implement a modified pairwise decomposition method for multilabel classification and compare its results to those of other approaches. We will also go deeper into the analysis of overlapping areas and how this information can be used to make optimizations.
Userfields: betreuer={ELM}
Authors Gasiorowski, Tomasz