Statistical Classification of Biosignature Information: Advancing Life Detection through Machine Learning

The search for extraterrestrial life is among the most compelling pursuits in planetary science, offering novel understandings of requirements for life, planetary exploration, and habitat identification. Central to this endeavor is the identification of biosignatures—measurable indicators of past or present life that are not formed by abiotic processes. The Statistical Classification of Biosignature Information (SCOBI) project is building and analyzing a multi-modal terrestrial analog dataset to create a binary classifier (a technique of machine learning) to accurately classify potential biosignature observations as indicative or non-indicative of life. This paper focuses on the creation of the standardized dataset for classifier training and evaluating feature importance across datatypes.