STARMIND: A FUZZY LOGIC KNOWLEDGE-BASED SYSTEM FOR THE AUTOMATED CLASSIFICATION OF STARS IN THE MK SYSTEM

Astrophysics is evolving toward a more rational use of costly observational data by intelligently exploiting the large terrestrial and spatial astronomical databases. In this paper, we present a study showing the suitability of an expert system to perform the classification of stellar spectra in the Morgan and Keenan (MK) system. Using the formalism of artificial intelligence for the development of such a system, we propose a rules' base that contains classification criteria and confidence grades, all integrated in an inference engine that emulates human reasoning by means of a hierarchical decision rules tree that also considers the uncertainty factors associated with rules. Our main objective is to illustrate the formulation and development of such a system for an astrophysical classification problem. An extensive spectral database of MK standard spectra has been collected and used as a reference to determine the spectral indexes that are suitable for classification in the MK system. It is shown that by considering 30 spectral indexes and associating them with uncertainty factors, we can find an accurate diagnose in MK types of a particular spectrum. The system was evaluated against the NOAO-INDO-US spectral catalog.