KIT | KIT-Bibliothek | Impressum | Datenschutz

An Experimental Evaluation of Time Series Classification Using Various Distance Measures

Górecki, Tomasz; Piasecki, Paweł

In recent years a vast number of distance measures for time series classification has been proposed. Obviously, the definition of a distance measure is crucial to further data mining tasks, thus there is a need to decide which measure should we choose for a particular dataset. The objective of this study is to provide a comprehensive comparison of 26 distance measures enriched with extensive statistical analysis. We compare different kinds of distance measures: shape-based, edit-based, feature-based and structure-based. Experimental results carried out on 34 benchmark datasets from UCR Time Series Classification Archive are provided. We use an one nearest neighbour (1NN) classifier to compare the efficiency of the examined measures. Computation times were taken into consideration as well.

Verlagsausgabe §
DOI: 10.5445/KSP/1000087327/07
Veröffentlicht am 22.11.2019
Cover der Publikation
Zugehörige Institution(en) am KIT Institut für Informationswirtschaft und Marketing (IISM)
Publikationstyp Zeitschriftenaufsatz
Publikationsjahr 2018
Sprache Englisch
Identifikator ISSN: 2363-9881
KITopen-ID: 1000100198
Erschienen in Archives of Data Science, Series A (Online First)
Band 5
Heft 1
Seiten A07, 25 S. online
KIT – Die Forschungsuniversität in der Helmholtz-Gemeinschaft
KITopen Landing Page