Anomaly Detection Using Explainable Random Forest for the Prediction of Undesirable Events in Oil Wells

Table 6

The final results obtained with the optimal hyper-parameters.

Dataset	Classifier	Accuracy (%)	Precision (%)	Recall (%)	F1-score (%)	AUC
Original dataset	LR	97.99	97.92	99.82	98.86	0.99
	RF	99.84	99.91	99.91	99.91	1
	DT	99.39	99.55	99.73	99.64	0.99
	K-NN	99.20	99.46	99.64	99.55	0.99
SMOTE dataset	LR	95.74	100	95.23	97.56	0.992
	RF	99.60	99.91	99.64	99.77	1
	DT	98.96	100	98.83	99.41	0.998
	K-NN	99.36	99.10	99.37	99.64	0.993