Detecting the Risk of Customer Churn in Telecom Sector: A Comparative Study

Table 5

Performance evaluation of applying HPO methods to the KNN classifier based on low-ratio undersampling strategy (0.3 and 0.4).

Dataset	Optimization algorithm	Accuracy	Recall	Precision	AUC	F1-score	MAE
Dataset 1	Default HPs	0.89	0.45	0.71	0.70	0.74	0.107
	GS	0.90	0.57	0.70	0.76	0.79	0.098
	RS	0.91	0.59	0.77	0.78	0.81	0.085
	GA-DEAP	0.91	0.58	0.73	0.77	0.80	0.092
	GA-TPOT	0.90	0.63	0.69	0.79	0.80	0.091
Dataset 2	Default HPs	0.75	0.46	0.56	0.66	0.67	0.253
	GS	0.80	0.60	0.66	0.73	0.74	0.202
	RS	0.80	0.64	0.65	0.75	0.75	0.199
	GA-DEAP	0.80	0.57	0.67	0.73	0.74	0.199
	GA-TPOT	0.80	0.57	0.67	0.73	0.74	0.199