Estimating the Mass of Galactic Components Using Machine Learning Algorithms

The estimation of galactic component masses can be carried out through various approaches that involve a host of assumptions about baryon dynamics or the dark matter model. In contrast, this work introduces an alternative method for predicting the masses of the disk, bulge, stellar, and total mass using the k-nearest neighbours, linear regression, random forest, and neural network (NN) algorithms, reducing the dependence on any particular hypothesis. The ugriz photometric system was selected as the set of input features, and the training was performed using spiral galaxies in Guo's mock catalogue from the Millennium simulation. In general, all of the algorithms provide good predictions for the galaxy's mass from 109 M circle dot to 1011 M circle dot, corresponding to the central region of the training domain. The NN algorithm showed the best performance. To validate the algorithm, we used the SDSS survey and found that the predictions of disk-dominant galaxies' masses lie within a 99% confidence level, while galaxies with larger bulges are predicted at a 95% confidence level. The NN also reveals scaling relations between mass components and magnitudes. However, predictions for less luminous galaxies are biased due to observational limitations. Our study demonstrates the efficacy of these methods with the potential for further enhancement through the addition of observational data or galactic dynamics.