CEESA meets machine learning: a Constant Elasticity Earth Similarity Approach to habitability and classification of exoplanets

Abstract

We examine the existing metrics of habitability and classification schemes of extrasolar planets and provide an exposition of the use of computational intelligence techniques to estimate habitability and to automate the process of classification of exoplanets. Exoplanetary habitability is a challenging problem in Astroinformatics, an emerging area in computational astronomy. The paper introduces a new constant elasticity habitability metric, the ‘Constant Elasticity Earth Similarity Approach (CEESA)’, to address the shortcoming of previous metrics. The proposed metric incorporates eccentricity as one of the component features to estimate the potential habitability of extrasolar planets. CEESA is a novel optimization model and computes habitability scores within the framework of a constrained optimization problem solved by metaheuristic method, mitigating the complexity and curvature violation issues in the process. The metaheuristic method, developed in the paper to solve the constrained optimization problem, is a ‘derivative-free’ optimization method, scope of which is promising beyond the current work. Habitabilty scores, such as CDHS (Bora et al., 2016), are recomputed with the imputed eccentricity values by the method developed in the paper and cross-matched with CEESA scores for validation. The paper also proposes fuzzy neural network-based approach to accomplish classification of exoplanets. Predicted class labels here are independent of CEESA, and are further validated by cross-matching them with the habitability scores computed by CEESA. We conclude by demonstrating the convergence between two proposed approaches, Earth-similarity approach (CEESA) and prediction of habitability labels (classification approach). The convergence between the two approaches establish the efficacy of CEESA in finding potentially habitable planets.