DaSSWeb - A combined distance measure between time series: experiments and applications
January 18th | 14:30
DaSSWeb, Data Science and Statistics Webinar
A combined distance measure between time series: experiments and applications
Margarida G. M. S. Cardoso
Associate Professor at the Department of Quantitative Methods for Management and Economics, Business School of ISCTE-University Institute of Lisbon
Associate researcher of BRU-Business Research Unit
Join us here.
The use of dissimilarity measures between time series is critical in several data analysis tasks ranging from simple querying to classification, clustering and anomaly detection. Recently, we proposed a new dissimilarity measure, COMB, a convex combination of four (normalized) distance measures between two time series.
In this webinar we first present some experiments resorting to the One Nearest Neighbor (1NN) classifier on labelled data to evaluate the comparative performance of COMB. We resort to data from the University of California Riverside (UCR) Time-Series Archive.
Two applications are also presented:
1. In order to better understand the degree of integration of the electricity markets of different European countries, we cluster time series regarding each country's day-ahead electricity market prices.
2. To reveal daily electricity consumption patterns, we cluster electricity consumption time series, data referring to the Portuguese Transmission System Operator (TSO).
Finally, some conclusions and further topics of research are discussed.
This work has been done in collaboration with Ana Martins and João Lagarto from ISEL-Instituto Superior de Engenharia de Lisboa.
Margarida G. M. S. Cardoso is an Associate Professor at the Department of Quantitative Methods for Management and Economics, Business School of ISCTE-University Institute of Lisbon. She is also an Associate researcher of BRU-Business Research Unit. Classification methods is the dominant topic of interest. Preferred areas of application are Marketing Research and Energy studies. More details can be found in: https://ciencia.iscte-iul.pt/authors/margarida-g-m-s-cardoso/cv