doi:10.6062/jcis.2015.06.02.0097 M´ odolo et al. 81 J. Comp. Int. Sci. (2015) 6(2):81-90 http://epacis.net/jcis/PDF_JCIS/JCIS-0097.pdf jcis@epacis.net ©2015 PACIS (http://epacis.net) PAN-AMERICAN ASSOCIACION OF COMPUTATIONAL INTERDISCIPLINARY SCIENCES An expert supernova spectral classification using artificial neural networks MarceloM´odolo a,c 1 , Lamartine N. F. Guimar˜aes b and Reinaldo R. Rosa a a Ph. D. Program in Applied Computer Science, National Institute for Space Research (INPE), S˜ ao Jos´ e dos Campos, SP, Brazil b Nuclear Energy Division, Institute for Advanced Studies, Department of Aerospace Science and Technology (IEAv/CTA), S˜ ao Jos´ e dos Campos, SP, Brazil c College of Engineerings, Technology and Information, Methodist University of S˜ ao Paulo (UMESP), S˜ ao Bernardo do Campo, SP, Brazil Received on june 14, 2015 / accepted on july 30, 2015. Abstract The supernovae (SN) classification is an important scientific issue in astrophysics and cosmology. Usu- ally, it is done by analyzing the spectra observed near the peak of the correspondent light curve. Because this task is difficult and usually made by an expert astronomer, it is important the study of computational techniques that allow the automatic classification of these spectra. In this paper we perform SN automatic classification method based on computational intelligence that simulates the human analytical expertise, making it a more formal classification and less prone to subjectivity of human analysis. Our classifier was developed using Multilayer Perceptron Neural Network to identify the usual SN types: Ia, Ib, Ic and II. The classifier was trained and tested on a database with 331 spectra of 56 different SN. The results are promising and indicate viability of this methodology for automatic SN classification in larger data sets. Keywords: Supernovae automatic classification, spectrum analysis, artificial neural network, computational data analysis. 1. Introduction The supernovae (SN) classification technique has been developed since 1941, recognizing two types of SN: Type I, characterized by hydrogen absence in its composition; and type II, with hydrogen presence [3]. The initial classification derived supernovae subtypes and the main current classification schemes consider nine different types of supernovae: Ia, Ib, Ic, IIb, IIL, IIP, IIF, IIn and IIpec. The subtypes determined by spectral properties are written in lower case and the subtypes determined by light curve properties are written in upper case. Usually, the SN spectrum analysis should be done soon after the explosion arises, close to its maximum light emission, which occurs about fifteen days after the star explosion. Figure 1 shows the basic SN GK classification scheme (after Giunt and Kim) [1] which identifies that type Ia are associated with thermonuclear explosion of white dwarf stars, while other types of SN are associated with core collapse of massive stars. 1 E-mail Corresponding Author: marmodolo@hotmail.com