Dark Matter Subhalo interpretations using machine learning: The fourth Fermi-LAT catalog

The quest for detecting dark-matter subhalos within the Galactic halo has taken many forms. Particularly interesting and promising is use of spectral degeneracies to distinguish otherwise indistinguishable gamma-ray sources with near-null star formation. In further exploration this realm, we attempt classify high-latitude, non-variable, unassociated Pulsar-like spectra in 20-70 GeV Dark Matter annihilation range. Implementing supervised machine learning models on 5788 recorded ten-year Fermi-LAT catalog (4FGL-DR2), where 1667 were formerly unassociated, a total 30 events over galactic latitude 10 degrees, |b| >= mean accuracy 98%. This classification allows us present subset potentially unanticipated as high-confidence Subhalo candidates.