Periodicity detection in AGN with the boosted tree method

TL;DR

This paper demonstrates the application of the XGBoost machine learning algorithm to detect periodicity in radio sources, offering a novel approach that complements traditional time series analysis methods.

Contribution

The study introduces a machine learning-based methodology using XGBoost for periodicity detection in radio astronomical data, providing an alternative to conventional techniques.

Findings

XGBoost effectively detects periodicity in radio source datasets.

The results align well with previous methods, validating the approach.

Machine learning offers new strategies for analyzing time series in astronomy.

Abstract

We apply a machine learning algorithm called XGBoost to explore the periodicity of two radio sources: PKS~1921-293 (OV~236) and PKS~2200+420 (BL~Lac), both radio frequency dataset obtained from University of Michigan Radio Astronomy Observatory (UMRAO), at 4.8 GHz, 8.0 GHz, and 14.5 GHz, between 1969 to 2012. From this methods, we find that the XGBoost provides the opportunity to use a machine learning based methodology on radio dataset and to extract information with strategies quite different from those traditionally used to treat time series and to obtain periodicity through the classification of recurrent events. The results were compared with other methods from others works that examined the same dataset and exhibit good agreement with them.