Abstract

As far as we know, galaxies form inside dark matter halos and elucidating this connection is a key element in theories of galaxy formation and evolution. In this work we propose a suite of machine learning tools to predict baryonic from halo properties in the IllustrisTNG300 magnetohydrodynamical simulation. We apply four methods: extremely randomized trees, K-nearest neighbors, light gradient boosting machine, and neural networks. Moreover, we combine the results of them in a stacked model. In addition, we apply all these methods in an augmented dataset using the synthetic minority over-sampling technique for regression with Gaussian noise, to deal with the problem of imbalanced data sets. Altogether, the ML algorithms are consistent at predicting central galaxy properties from a set of input halo properties such as halo mass, concentration, spin, and halo overdensity. For stellar mass, the Pearson correlation coefficient is 0.98, while for specific star formation rate, color, and size it is between 0.7–0.8. Lastly, the presented analysis adds evidence to previous works indicating that certain galaxy properties cannot be reproduced using halo features alone.