Selection of powerful radio galaxies with machine learning

Supplementary material to the article "Selection of powerful radio galaxies with machine learning" from Carvajal et al., 2023. Preprint version can be obtained from https://arxiv.org/abs/2309.11652

Included files are:

• predicted_rAGN_HETDEX.parquet: Dataset from HETDEX Spring field. It includes initial properties from sources as well as predicted values. Description of columns in Appendix G of Carvajal et al., 2023.
• predicted_rAGN_S82.parquet: Dataset from Stripe 82 field. It includes initial properties from sources as well as predicted values. Description of columns in Appendix G of Carvajal et al., 2023.
• classification_AGN_galaxy.pkl: Model for classification between AGN and galaxies. It takes as input features described in article. It delivers uncalibrated scores.
• classification_radio_detection.pkl: Model for classification between radio detectable and radio non-detectable AGN. It takes as input features described in article. It delivers uncalibrated scores.
• regression_redshift.pkl: Model for prediction of redshift for radio-detectable AGN. It takes as input features described in article.
• cal_classification_AGN_galaxy.joblib: Calibrated model for classification between AGN and galaxies. It takes as input scores from uncalibrated model. It delivers calibrated probabilities.
• cal_classification_radio_detection.joblib: Calibrated model for classification between radio detectable and radio non-detectable AGN. It takes as input scores from uncalibrated model. It delivers calibrated probabilities.
• datasets_description.txt: Description of columns in parquet files.

Files with parquet extension were generated with python using the pandas package (v.1.4.2) and the engine fastparquet.

Files with pkl extension were generated with python using pycaret (v.2.3.10).

Files with joblib extension were generated with python using scikit-learn (v.0.23.2).

An example on how to use these files can be found in https://github.com/racarvajal/ML_prediction_pipeline_run