Exploring galactic properties with machine learning Predicting star formation, stellar mass, and metallicity from photometric data

TL;DR

This paper demonstrates that machine learning models, specifically CatBoost and deep learning, can accurately predict key galaxy properties such as star formation rate, stellar mass, and metallicity from photometric data, aiding large-scale astronomical analysis.

Contribution

The study introduces a machine learning framework that effectively predicts galaxy properties from photometric data, achieving low errors and showcasing potential for automated galaxy analysis.

Findings

Achieved RMSE of 0.336 dex for star formation rate

Reduced stellar mass prediction error to 0.206 dex

Predicted metallicity with RMSE of 0.097 dex

Abstract

Aims. We explore machine learning techniques to forecast star formation rate, stellar mass, and metallicity across galaxies with redshifts ranging from 0.01 to 0.3. Methods. Leveraging CatBoost and deep learning architectures, we utilize multiband optical and infrared photometric data from SDSS and AllWISE, trained on the SDSS MPA-JHU DR8 catalogue. Results. Our study demonstrates the potential of machine learning in accurately predicting galaxy properties solely from photometric data. We achieve minimised root mean square errors, specifically employing the CatBoost model. For star formation rate prediction, we attain a value of RMSESFR = 0.336 dex, while for stellar mass prediction, the error is reduced to RMSESM = 0.206 dex. Additionally, our model yields a metallicity prediction of RMSEmetallicity = 0.097 dex. Conclusions. These findings underscore the significance of automated methodologies in efficiently estimating critical galaxy properties, amid the exponential growth of multi-wavelength astronomy data. Future research may focus on refining machine learning models and expanding datasets for even more accurate predictions.