G-mode Clustering Method applied to Asteroid Taxonomy
Authors: Hasselmann, P. H., Carvano, J. M., Lazzaro, D.
G-mode is a clustering method developed by A. I. Gavrishin in the 60's for geochemical classification of rocks, but was also applied to asteroid photometry, cosmic rays, lunar sample and planetary science spectroscopy data. In this work, we used it to classify the asteroids from SDSS Moving Objects Catalog. The method works identifying normal distributions in a multidimensional space of variables. The identification starts with finding the closest points in the sample, which is a consumable problem when the data is not planar. Therefore, to achieve satisfying results in a human bearable time, we implemented the method, which was previously written in FORTRAN 77, in PYTHON 2.7.2 and using Numpy, Scipy and Matplotlib packages. The Numpy was used for array and matrix manipulation and Matplotlib for plot control. The Scipy had a import role in speeding up G-mode, Scipy.cKD-Tree and Numpy.histogramdd were applied to find the initial seeds from which the clusters are going to evolve. Scipy was also used to quickly produce dendograms showing the distances between the clusters.
Finally, we will present our results for Asteroids Taxonomy and tests of the our code for different sample sizes and implementations.