Functional Principal Components direction to cluster earthquake

Vienna (Austria)

Looking for curves similarity could be a complex issue characterized by subjective choices related to continuous
transformations of observed discrete data (Chiodi, 1989).
In this paper we combine the aim of finding clusters from a set of individual curves to the functional nature of data,
applying a variant of a k-means algorithm based on the principal component rotation of data. We apply a classical
clustering method to rotated data, according to the direction of maximum variance.
A k-means clustering algorithm based on PCA rotation of data is proposed, as an alternative to methods that
require previous interpolation of data based on splines or linear fitting (García-Escudero and Gordaliza (2005),
Tarpey (2007), Sangalli et al. (2008)).

[1] Chiodi, M. (1989). The clustering of longitudinal data when time series are short. Multivariate data analysis,
pages 445–453.
[2] García-Escudero, L. A. and Gordaliza, A. (2005). A proposal for robust curve clustering. Journal of classification,
22, 185–201.
[3] Gillard, D., R. A. and Okubom, P. (1996). Highly concentrated seismicity caused by deformation of kilauea’s
depp magma system. Nature, 384, 343–346.
[4] Hastie, T. J. (1997). Generalized additive models. Chapman and Hall, London.
[5] Menke, W. (1999). Using waveform similarity to constrain earthquake locations. Bull.Seismol. Soc. Am., 89,
1143–1146.
[6] Mezcua, J. and Rueda, J. (1994). Earthquake relative location based on waveform similarity. Tectonophysics,
233, 253–263.
[7] Phillips, W. S., H. L. F. J. (1997). Detailed joint structure in a geothermal reservoir from studies of induced
microearthquake studies. Journal of Geophysical Research, 102, 745–763.
[8] Ramsey, J. O. and Silverman, B. W. (2006). Functional Data Analysis. Springer, New York.
[9] Sangalli, L. M., Secchi, P., Vantini, S., and Vitelli, V. (2008). K-means alignment for curve clustering. MOX
(Modeling and Scientific Computing)-Report, 13.
[10] Tarpey, T. (2007). Linear transformations and the k-means clustering alghoritm: applications to clustering
curves. American statistician, 61(1), 34–40.