A Graph-Driven Approach to Complex Challenges: A Case Study on Multiobjective Stellar and Earth-Like Exoplanet Clustering

Resumo

This study investigates the complex interplay between stellarmetallicity, stellar mass, and the likelihood of hosting Earth-like plan-ets. Recognizing that the formation of habitable planets is a multifac-torial process, we adopt a novel graph-based approach that integratesLocal-Sensitive Hashing (LSH) for efficient dimensionality reduction withmodularity-based clustering to analyze a dataset of nearly 38,000 exo-planets. Our methodology enables us to identify distinct planetary com-munities and examine key host star properties. Comprehensive statis-tical analyses, including ANOVA, t-tests, and various correlation mea-sures (Pearson, Spearman, Kendall, distance correlation, and mutual in-formation), were employed to explore the relationships between stellarmass and planetary density. While our findings indicate that Sun-likestars—particularly those with slightly lower stellar mass—are more likelyto host Earth-like planets, the observed correlations suggest that the in-fluence of stellar mass is relatively weak, pointing to the involvement ofadditional factors in planetary formation. Moreover, the computationalefficiency of our approach highlights its potential applicability in otherfields that deal with high-dimensional, complex datasets, such as sys-tems biology and social network analysis. Overall, our work contributesto a deeper understanding of the determinants of planetary habitabilityand underscores the need for more sophisticated multivariate models tocapture the full complexity of planet formation.