An automated activity classification tool for optical galaxy spectra

¹ Physics Department, and Institute of Theoretical and Computational Physics, University of Crete, 71003 Heraklion, Greece
² Institute of Astrophysics, Foundation for Research and Technology-Hellas, 71110 Heraklion, Greece
³ Center for Astrophysics | Harvard & Smithsonian, 60 Garden St., Cambridge, MA 02138, USA
⁴ Astronomical Institute, Academy of Sciences, Boční II 1401, CZ-14131 Prague, Czech Republic
⁵ ALMA Sistemi Srl, Guidonia, Rome 00012, Italy
⁶ Quantum Innovation Pc, Chania 73100, Greece

^★ Corresponding author: This email address is being protected from spambots. You need JavaScript enabled to view it.

Received: 10 July 2025
Accepted: 23 February 2026

Abstract

Context. Reliable and versatile galaxy activity diagnostic tools are indispensable for comprehending the physical processes that drive galaxy evolution. Traditional methodologies frequently necessitate extensive preprocessing, such as starlight subtraction and emission-line deblending (e.g., Hα and [N II]), which can introduce substantial biases and uncertainties due to their model-dependent nature. Additionally, numerous diagnostics omit the inclusion of dormant (passive) galaxies.

Aims. This work aims to develop a reliable, automated, and efficient diagnostic tool capable of distinguishing among star-forming galaxies, active galactic nuclei (AGNs), low-ionization nuclear emission-line regions (LINERs), composite and passive galaxies under one unified scheme.

Methods. We developed a diagnostic tool based on a support vector machine trained on ground-truth data originating from optical emission-line ratios and color selection criteria. Building upon previous literature findings and exploring various combinations of discriminatory feature schemes, we identified the equivalent widths (EWs) of Hβ, [O III] λ5007, and Hα+[N II] λλ6548,84 as key discriminatory features. Additionally, galaxies classified as AGNs can be distinguished as broad- or narrow-line AGNs by measuring the full quarter at the half-maximum of the Hα and [N II] complex.

Results. Employing machine-learning algorithms and three EWs directly measured from the galaxy’s optical spectrum, we have developed a diagnostic tool that encompasses all potential activities of galaxies while simultaneously achieving high-performance scores across all of them. Our diagnostic achieves overall accuracy of ∼83% and recall of ∼79% for star-forming galaxies, ∼94% for AGN, ∼85% for LINERs, ∼77% for composite galaxies, and ∼96% for passive galaxies.

Conclusions. Our diagnostic tool offers significant improvements over the existing galaxy activity diagnostics as it can be applied to large numbers of spectra, eliminates the need for preprocessing (i.e., starlight subtraction or flux calibration) and spectral-line deblending, encompasses all activity classes under one unified scheme, and offers the ability to distinguish between the two main types of AGN. In addition, the omission of starlight subtraction was not found to significantly reduce the diagnostic’s performance. Furthermore, the narrow wavelength range required for its application enables its use over a wide range of redshifts, making it highly relevant to activity studies of high-redshift galaxies.