Clustering and Classification of Kepler's Confirmed Exoplanets Based on Mixture Models

摘要

Exoplanets are planets which orbit a star beyond our solar system. NASA's space based Kepler telescope has discovered thousands of confirmed exoplanet candidates using planetary transit method. Various types of planets with different physical characteristics are present in this Keplars' confirmed exoplanet group. We split this confirmed exoplanet group into three different temperature groups viz. hot, warm, and comfort according to effective equilibrium temperature. Using different clustering validity indices we try to find the optimum number of homogeneous groups in the radius-mass, and mass-density spaces of different exoplanet groups individually. We split each planetary group into optimum number of homogeneous subpopulations according to optimum clustering technique. For classification of habitable exoplanets, we try to fit different probability distributions on the homogeneous subpopuations of habitable exoplanets with four physical parameters and the best fitted distribution has been chosen according to information criterion values. We propose a classification rule based on conditional probability of class membership which depends on mixture of different distributions. We try to classify few habitable exoplanets using this rule and compare our results with the true class of these planets.