Evaluating Clustering Algorithms as Applied to Exoplanet Populations

Abstract

With the rapid rate of exoplanet discoveries, established exoplanet populations need to be continually revisited and updated. Identifying clusters of exoplanets based on major characteristics is a data- driven way to accomplish this. In this paper, I cluster exoplanets based on planetary mass, radius, insolation flux, and multiplicity, as well as stellar age, metallicity, multiplicity, and gravitational constant. However, examining clusters identified by single clustering algorithms does not provide information on how robust these clusters are. This paper presents an algorithm to match similar clusters found across different clustering algorithms and subsets of data. By comparing clusters based on how consistently they are identified across iterations of clustering, I determine the most robust patterns and evaluate which clustering algorithms are the most effective at identifying robust clusters.