In this work we present the Gaussian Probabilistic Classification Method (GPCM) as a Machine Learning classifier applied to the Gaia photometric science alerts data (GPSA). We employ a k-d tree cross-matcher with the full third data release from Gaia satellite in combination with the GPSA data-base to extract magnitude and color data. Selection cuts in full Gaia data release are given by selecting sources above the galactic plane in the interval (70◦ < b < 90◦) with magnitude given by 14 ≤ M ≤ 20. We apply the method for classifying unknown sources inside the GPSA data by using already tagged objects inside the following classes: Supernovae type Ia (SN Ia), Supernovae type II (SN II), quasi-stellar objects (QSO) and active galactic nuclei (AGN). During the training of the classifier we find the the Tied method gives the highest accuracy with 80% of correctly classified objects. By applying the trained classifier we find that most of the unknown sources in the GPSA catalog are tagged as Quasi-stellar objects (QSO) with 52% of total the predictions falling within this category for the unknown GPSA sources.
echalbau / ml_gaia Goto Github PK
View Code? Open in Web Editor NEWLicense: MIT License