Identifying radio active galactic nuclei with machine learning and large-area surveys

Home

All issues

Volume 701 (September 2025)

A&A, 701 (2025) A179

Full HTML

Open Access

Table 1.

Hyperparameters and model performance for different machine learning algorithms.

Algorithm	Hyperparameters	AGN			non-AGN
Algorithm	Hyperparameters	precision	recall	F₁ score	precision	recall	F₁ score
SVC	C = 1.0, kernel = ′rbf′,	0.976	0.718	0.827	0.875	0.991	0.929
SVC	gamma = ′scale′	0.976	0.718	0.827	0.875	0.991	0.929
MLP	solver = ′adam′, alpha = 0.0001,	0.959	0.829	0.890	0.920	0.982	0.950
MLP	hidden_layer_sizes = (100, )	0.959	0.829	0.890	0.920	0.982	0.950
RF	n_estimators = 100, max_features = ′sqrt′,	0.986	0.847	0.911	0.928	0.994	0.960
	max_samples = None, min_samples_split = 2,
	bootstrap = True, max_leaf_nodes = None,
	max_depth = None
AdaBoost	n_estimators = 50, learning_rate = 1.0,	0.947	0.847	0.894	0.927	0.976	0.951
AdaBoost	estimator = None	0.947	0.847	0.894	0.927	0.976	0.951
HGBT	max_iter = 100, max_leaf_nodes = 31,	0.973	0.853	0.909	0.931	0.988	0.959
	max_depth = None, min_samples_leaf = 20,
	max_bins = 255, l2_refularization = 0.0,
	early_stopping = ′auto′
CatBoost (default)	iterations = 1000	0.974	0.865	0.916	0.936	0.988	0.961
CatBoost (default)	learning_rate = 0.01108, depth = 6	0.974	0.865	0.916	0.936	0.988	0.961
CatBoost (optimal)	iterations = 184 (10, 2000)*	0.981	0.888	0.932	0.946	0.991	0.968
	learning_rate = 0.317 (0.001, 1)*,
	depth = 6 (1, 16)*

Notes. The lower and upper limits for hyperparameter tuning are presented in brackets.

Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.

Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.

Initial download of the metrics may take a while.