Page 241 - Special Topic Session (STS) - Volume 1

STS426 Asis K.C.

Clustering and classification of Astronomical
objects- A new paradigm in Statistics
Asis Kumar Chattopadhyay*
Department of Statistics, University of Calcutta, India
Visiting Scholar, Concordia University, Canada

Abstract
This collection of works involves the application of statistics to astronomy and
the development of statistical methods to solve the problems related to the
universe, leading us to discoveries of new astrophysical phenomena. Data
collection missions like Galaxy Evolution Explorer, Kepler Space Telescope,
Hubble Space Telescope and virtual archives like Sloan Digital Sky Survey,
Multi-mission Archive at STSCI, NASA Extragalactic Data base preserve
petabytes of data, which can be used for big data analyses. Usually collection
of data on celestial bodies is obscured by bad weather conditions, obstruction
by another celestial object or instrumental restrictions and it cannot be
repeated. Hence we often get data contaminated with noise, affected by
outliers or sparsely distributed. In all such situations, the usual statistical
methods fail and we need to use their adaption or to introduce new methods
as per requirements. To overcome such problems, there are various
transformations and denoising techniques available in the literature (e.g.
kernel principal component analysis (KPCA)). Sometimes there are rare objects,
unevenly spaced data of unequal lengths where classical statistical methods
are only applicable when the data is interpolated to get into a form suitable
for the methods. We have suggested some possible solutions under the above
scenario.

Keywords
Astrostatistics, Clustering, Classification, Kernel Principal Component Analysis,
Missing Value.

1. Introduction
Astronomy, perhaps the oldest observational science, has got its
spectacular emergence with the advent of theoretical astrophysics. During the
last two decades galaxy formation theory and their related star formation
histories have drawn interests among the astrophysicists to a great extent to
uncover these mysteries using the reach treasure of virtual archives. Scientists
working in the theoretical areas like cosmology and relativity are gradually
becoming interested in database analysis because of technological
advancements enabling us to have data related to such physical phenomenon.
While digging the pathway, a new branch (Chattopadhyay et al.(2014))
230 | I S I W S C 2 0 1 9