Classification of Fermi-LAT unidentified gamma-ray sources using catboost gradient boosting decision trees

Abstract

The latest Fermi-LAT gamma-ray catalogue, 4FGL-DR3, presents a large fraction of sources without clear association to known counterparts, i.e. unidentified sources (unIDs). In this paper, we aim to classify them using machine learning algorithms, which are trained with the spectral characteristics of associated sources to predict the class of the unID population. With the state-of-the-art catboost algorithm, based on gradient boosting decision trees, we are able to reach a 67 per cent accuracy on a 23-class data set. Removing a single of these classes – blazars of uncertain type – increases the accuracy to 81 per cent. If interested only in a binary AGN/pulsar distinction, the model accuracy is boosted up to 99 per cent. Additionally, we perform an unsupervised search among both known and unID population, and try to predict the number of clusters of similar sources, without prior knowledge of their classes. The full code used to perform all calculations is provided as an interactive python notebook.

Classification of Fermi BCUs Using Machine Learning

“…We use KNN (Arthur & Vassilvitskii 2006;Xu et al 2020), which determines the category of the sample to be divided according to the category of the nearest one or several samples. We use gradient boosting + categorical features (CB; Prokhorenkova et al 2017;Coronado-Blázquez 2022, which supports categorical variables and high accuracy gradient boosting decision tree (GBDT) framework. LR aims to map the results of linear regression to the interval from 0 to 1 through the logistic function, and the classification of the data is obtained by comparing with 0.5.…”

“…Supervised machine learning (SML) is a useful and alternative classification method and it provide reference for classification results. SML had been used by many scholars (Ackermann et al 2012;Chiaro et al 2016;Saz Parkinson et al 2016;Salvetti et al 2017;Lefaucheur & Pita 2017;Yi et al 2017;Kovačević et al 2019Kovačević et al , 2020Kang et al 2019a;Xu et al 2020;Xiao et al 2020;Zhu et al 2021b;Coronado-Blázquez 2022) to classify BCUs from the Fermi-LAT catalogs.…”

A study of algorithms for solving nonlinear two-level programming problems oriented to decision tree models

“…Decision trees are a top-down, divide-and-conquer recursive process, where the comparison of attributes is also the comparison of node attribute values within the decision tree [11][12]. Each path from the root node to the leaf nodes forms a classification rule that recurses to the leaf nodes to obtain a conclusion.…”

“…Jinshan Lin, Min Lin and Hang Xu. Applied Mathematics and Nonlinear Sciences, 9(1) (2024)[1][2][3][4][5][6][7][8][9][10][11][12][13][14][15][16][17] …”

Identifying the Physical Origin of Gamma-Ray Bursts with Supervised Machine Learning

“…However, the abovementioned studies predominantly use machine-learning methods of the unsupervised type, where only the observed features of the GRBs are inputted into the models, but not the labels (the GRBs' physical classes being Type I or II). On the other hand, the other type of machinelearning methods, supervised methods, are also commonly employed by astronomy researchers in the classification of other astronomical objects (e.g., Luo et al 2023;Zhu-Ge et al 2023;Connor & van Leeuwen 2018;Butter et al 2022;Coronado-Blázquez 2022;de Beurs et al 2022;Fan et al 2022;Villa-Ortega et al 2022;Yang et al 2022a;Kaur et al 2023), although studies on the application of supervised methods on GRB are scarce. Since supervised methods take both features and labels as input, and can produce deterministic predictions of the class of new GRBs, they can be helpful in identifying the true physical origin of intermingled GRBs.…”