Multiscale Cluster Analysis
Bjørn Kåre Alsberg · Analytical Chemistry · 1999
This article describes how the concept of multiresolution is used with cluster analysis of spectral data. Multiresolution analysis progressively increases the resolution of a spectrum profile by adding levels of details contained in scales obtained from a discrete wavelet transform. At each resolution level a cluster analysis is performed on all the spectra profiles in the data set. This allows the relating of changes in the cluster pattern to various broad and narrow features in the spectral data profiles. The analysis also provides an approximate location of the important features in the original wavenumber domain.