A statistical study of lopsided galaxies using random forests

Context. Lopsided galaxies are late-type galaxies that feature a non-axisymmetric disk caused by an uneven distribution of their stellar mass or light. Despite being a relatively common perturbation, there are still several questions regarding its origin and the information that can be extracted from them about the evolutionary history of late-type galaxies. Previous observational and numerical studies have suggested a correlation between lopsidedness and galaxy assembly history and internal structure. Aims. The advent of several large multiband photometric surveys will allow us to statistically analyze this perturbation, with information that was not previously available. This paper aims to develop a method of rapidly and automatically pre-classifying late-type galaxies as lopsided and symmetric, purely based on the galaxies’ internal parameters. This method allows us to test the hypothesis that lopsidedness is a strong indicator of peculiar internal galaxy properties, rather than an indicator of the present-day environment they are hosted in. Methods. We selected a sample of ≈8000 late type galaxies from the Illustris TNG50 simulation. A Fourier decomposition of their stellar mass surface density was used to label galaxies as lopsided and symmetric. We trained a random forest binary classifier to rapidly and automatically identify this type of perturbation, exclusively using galaxies’ internal properties. We explored different algorithms to deal with the imbalance nature of our data, and selected the most suitable approach based on the considered metrics. Results. We show that our trained algorithm can provide a very accurate and rapid pre-classification of lopsided galaxies. The excellent results obtained by our classifier, trained with features that do not account for the galaxies environment, strongly support the hypothesis that lopsidedness is mainly a tracer of galaxies’ internal structures. We also show that similar results can be obtained when considering as input features observable quantities that are readily obtainable from multiband photometric surveys. Conclusions. Our results show that algorithms such as the ones considered allow for the rapid and accurate pre-classification of lopsided galaxies from large multiband photometric surveys, allowing us to explore whether lopsidedness in present-day disk galaxies is connected to galaxies’ specific evolutionary histories.