CRND: An Unsupervised Learning Method to Detect Network Anomaly

Table 4

The evaluation results during linear classification protocol.

Number of five-tuple	Accuracy (%)	Precision (%)	Recall (%)	False-alarm (%)	F1-score
10	82.35	89.67	68.84	6.53	0.78
50	90.49	98.17	90.46	1.24	0.88
100	88.85	98.73	76.29	0.81	0.86
10,000	91.94	98.93	83.03	0.72	0.91