TY  - CONF
ID  - moewes_adjustung_2008
T1  - Adjusting Monitored Experiments to Real-World Cases by Matching Labeled Time Series Motifs
A1  - Moewes, Christian
A1  - Kruse, Rudolf
ED  - Mikut, Ralf
ED  - Reischl, Markus
TI  - Proceedings 18. Workshop Computational Intelligence, Dortmund, 3. - 5. Dezember 2008
T3  - Schriftenreihe des IAI, Universität Karlsruhe (TH)
Y1  - 2008
SP  - 214
EP  - 223
PB  - Universitätsverlag Karlsruhe
AD  - Karlsruhe, Germany
SN  - 978-3-86644-282-5
UR  - http://digbib.ubka.uni-karlsruhe.de/volltexte/1000009271
M2  - doi: 10.5445/KSP/1000009271
KW  - Frequent Pattern Mining
KW  - Labeling
KW  - Motif Discovery
KW  - Multivariate Time Series Analysis
N2  - In this paper we devote ourselves to the difficulty of fitting human designed experiments to real-world cases. We decompose this problem into two smaller subproblems: 1.) The search of recurrent patterns in temporal sequences, so called motifs that are deemed to be discovered in both the experiments and the real observations and 2.) the matching of motifs to linguistic terms which are possibly available as domain knowledge. Therefore we describe an effective time series representation that enormously speeds up the search for these motifs. We present some approaches to adjust the designed experiments with the help of the discovered motifs. Finally, we conclude our work and give prospects to possible extensions.
ER  -