Online Program

Friday, February 20
PS2 Poster Session 2 & Refreshments Fri, Feb 20, 5:15 PM - 6:30 PM
Napoleon AB

Classification of Time Series Using Similarity Analysis (303009)

*David J. Corliss, Wayne State University, Ford Motor Company 

Keywords: Time Series Analysis, Similarity Analysis, Astrostatistics

Similarity analysis computes a variety of difference measures between sets of sample data. These measures are used to create a similarity matrix, to which further analysis can be applied. In the case of time-ordered data, this matrix measures the degree to which two events occurring over a period of time follow the same pattern. This method is illustrated using the classification of light curves: the observed brightness of an astronomical object as a function of time. The data consist of light curves from 300 supernova events captured by the Sloan Digital Sky Survey. Similarity matrices serve as input to cluster analysis, identifying groups of supernova events with similar properties. An additional example is given, applying this method to classify econometric events.