Christian D. Klose, Affiliation: Think GeoHazards
Clustering Approach for Partitioning Directional Data in Earth and
Space Sciences
A simple clustering approach, based on vector quantization (VQ), is
presented for partitioning directional data in Earth and Space
Sciences. Directional data are grouped into a certain number of
disjoint isotropic clusters and, at the same time, the average
direction is calculated for each group. The algorithm is fast and,
thus, can be easily utilized for large data sets. It shows good
clustering results compared to other benchmark counting methods for
directional data. No heuristics is being used, because the grouping
of data points, the binary assignment of new data points to clusters,
and the calculation of the average cluster values are based on the
same cost function.