Classificação de exoplanetas rochosos utilizando inteligência artificial: uma abordagem de iniciação científica no Ensino Médio

Abstract

The search for habitable planets is one of the greatest challenges in contemporary astronomy. However, this search is full of controversies, difficulties, and, so far, remains inconclusive, mainly due to the lack of direct observations of these worlds. In light of these limitations, it becomes necessary to develop alternative methods to identify exoplanets with potential habitability. We therefore propose a classification of rocky exoplanets based on their physical, chemical, and orbital parameters, using unsupervised learning algorithms to identify analogs to planets in the Solar System. Data were obtained from the NASA Exoplanet Archive and Exoplanet.eu catalogs, followed by a cleaning and preprocessing phase. The Agglomerative Hierarchical Clustering technique was applied and validated using Elbow, Silhouette, Calinski Harabasz, and Davies-Bouldin metrics. The results revealed 340 Mercury analogs, 44 Venus analogs, 2 Earth analogs, and no Mars analogs. This study also explores how the resulting data can be used in educational settings, supporting the development of an Astronomy elective and encouraging critical thinking in the face of misinformation. Finally, the identified clusters may assist in selecting promising targets for future exoplanet characterization missions.