A data mining approach to recognizing source classes for unassociated gamma-ray sources

Research output: Contribution to journalConference article

Abstract

The Fermi-LAT 3rd source catalog (3FGL) provides the gamma-ray properties for 3034 gamma- ray sources. While 2024 sources in the 3FGL are associated with AGNs (58 % of the total), pulsars (5 %) and the other classes (4 %), 1010 sources (33 %) remain as unassociated sources. In recognizing source classes for unassociated gamma-ray sources of the Fermi-LAT source cat- alogs, various data mining techniques have been applied, e.g. classification tree and artificial neural network. As a robust alternative to these data mining techniques, we present the Maha- lanobis Taguchi (MT) method to recognize source classes. The MT method creates a multidimen- sional unit space from characteristic variables of a normal class (e.g. AGN) to identify sources of the normal class from those of the other classes using Mahalanobis distances. In this paper, we present the results of the source classification for the unassociated gamma-ray sources in 3FGL by applying the MT method. We also discuss a possibility of dark matter Galactic subhalos for the unclassified sources at jbj > 20°.

Original languageEnglish
Article number857
JournalProc. of 34th ICRC
Volume30-July-2015
Publication statusPublished - 2015

Fingerprint

data mining
Taguchi methods
catalogs
pulsars
dark matter
gamma rays

ASJC Scopus subject areas

  • General

Cite this

A data mining approach to recognizing source classes for unassociated gamma-ray sources. / Yoshida, Kenji.

In: Proc. of 34th ICRC, Vol. 30-July-2015, 857, 2015.

Research output: Contribution to journalConference article

@article{71de5a32f43947c7955c14a76ffd930e,
title = "A data mining approach to recognizing source classes for unassociated gamma-ray sources",
abstract = "The Fermi-LAT 3rd source catalog (3FGL) provides the gamma-ray properties for 3034 gamma- ray sources. While 2024 sources in the 3FGL are associated with AGNs (58 {\%} of the total), pulsars (5 {\%}) and the other classes (4 {\%}), 1010 sources (33 {\%}) remain as unassociated sources. In recognizing source classes for unassociated gamma-ray sources of the Fermi-LAT source cat- alogs, various data mining techniques have been applied, e.g. classification tree and artificial neural network. As a robust alternative to these data mining techniques, we present the Maha- lanobis Taguchi (MT) method to recognize source classes. The MT method creates a multidimen- sional unit space from characteristic variables of a normal class (e.g. AGN) to identify sources of the normal class from those of the other classes using Mahalanobis distances. In this paper, we present the results of the source classification for the unassociated gamma-ray sources in 3FGL by applying the MT method. We also discuss a possibility of dark matter Galactic subhalos for the unclassified sources at jbj > 20°.",
author = "Kenji Yoshida",
year = "2015",
language = "English",
volume = "30-July-2015",
journal = "Proc. of 34th ICRC",

}

TY - JOUR

T1 - A data mining approach to recognizing source classes for unassociated gamma-ray sources

AU - Yoshida, Kenji

PY - 2015

Y1 - 2015

N2 - The Fermi-LAT 3rd source catalog (3FGL) provides the gamma-ray properties for 3034 gamma- ray sources. While 2024 sources in the 3FGL are associated with AGNs (58 % of the total), pulsars (5 %) and the other classes (4 %), 1010 sources (33 %) remain as unassociated sources. In recognizing source classes for unassociated gamma-ray sources of the Fermi-LAT source cat- alogs, various data mining techniques have been applied, e.g. classification tree and artificial neural network. As a robust alternative to these data mining techniques, we present the Maha- lanobis Taguchi (MT) method to recognize source classes. The MT method creates a multidimen- sional unit space from characteristic variables of a normal class (e.g. AGN) to identify sources of the normal class from those of the other classes using Mahalanobis distances. In this paper, we present the results of the source classification for the unassociated gamma-ray sources in 3FGL by applying the MT method. We also discuss a possibility of dark matter Galactic subhalos for the unclassified sources at jbj > 20°.

AB - The Fermi-LAT 3rd source catalog (3FGL) provides the gamma-ray properties for 3034 gamma- ray sources. While 2024 sources in the 3FGL are associated with AGNs (58 % of the total), pulsars (5 %) and the other classes (4 %), 1010 sources (33 %) remain as unassociated sources. In recognizing source classes for unassociated gamma-ray sources of the Fermi-LAT source cat- alogs, various data mining techniques have been applied, e.g. classification tree and artificial neural network. As a robust alternative to these data mining techniques, we present the Maha- lanobis Taguchi (MT) method to recognize source classes. The MT method creates a multidimen- sional unit space from characteristic variables of a normal class (e.g. AGN) to identify sources of the normal class from those of the other classes using Mahalanobis distances. In this paper, we present the results of the source classification for the unassociated gamma-ray sources in 3FGL by applying the MT method. We also discuss a possibility of dark matter Galactic subhalos for the unclassified sources at jbj > 20°.

UR - http://www.scopus.com/inward/record.url?scp=84988734740&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84988734740&partnerID=8YFLogxK

M3 - Conference article

AN - SCOPUS:84988734740

VL - 30-July-2015

JO - Proc. of 34th ICRC

JF - Proc. of 34th ICRC

M1 - 857

ER -