Table 6. Classification performances on bot detection.

Approach	Hyper-parameters		Performances
	Values explored via grid search	Best values	Acc.	F1	ROC-AUC	Precision	Recall
Decision Tree	Criterion (gini/entropy),	entropy	0.885	0.903	0.890	0.938	0.871
	max_depth (1, 2,…, 10),	5
	min samples leaf (2, 3, …, 20),	14
	max leaf nodes (1, 2, …, 20)	17
eXtreme Gradient Boosting (XGBoost) for trees	Max depth (5, 6, …, 10),	5	0.892	0.909	0.893	0.933	0.887
	alpha (0.1, 0.3, 0.5),	0.5
	learning rate (0.01, 0.02, …, 0.05),	0.02
	estimators (100, 200, 300)	200
Random Forests	Max depth (5, 6, …, 10),	8	0.898	0.915	0.899	0.934	0.897
	criterion (gini/entropy),	entropy
	estimators (100, 200, 300)	200