Attack detection in water distribution systems using machine learning

Table 3 A comparison of the algorithms on noisy BATADAL dataset

Rank	Name	No. attacks	S	\(S_{TTD}\)	\(S_{CLF}\)	\(F_1\)	TPR	TNR	PPV	TP	FP	TN	FN
1	QDA	7	0.9441	0.9584	0.9298	0.8870	0.8870	0.9727	0.8870	361	46	1636	46
2	MD	7	0.9010	0.9024	0.8995	0.8719	0.8108	0.9881	0.9429	330	20	1662	77
3	Ensemble	7	0.8609	0.9040	0.8178	0.6802	0.7420	0.8936	0.6279	302	179	1503	105
4	LOF	7	0.8507	0.8814	0.8201	0.7208	0.6978	0.9423	0.7454	284	97	1585	123
5	SOD	6	0.8267	0.8172	0.8362	0.6660	0.8354	0.8371	0.5537	340	274	1408	67
6	Naive	7	0.7500	1.0000	0.5000	0.3261	1.0000	0.0000	0.1948	407	1682	0	0
7	OSVM	6	0.6420	0.6449	0.6392	0.4351	0.2801	0.9982	0.9744	114	3	1679	293
8	LDA	4	0.5117	0.4444	0.5790	0.2737	0.1597	0.9982	0.9559	65	3	1679	342