Anomaly detection and community detection in networks

Journal of Big Data

Table 5 The confusion matrix for the network of POLBLOGS with injected edges (Experiment 1)

	\(\pi\)	Precision	Recall	F1
High recall/precision	0.3	0.49	0.99	0.65
High precision	0.5	0.45	0.27	0.34
High recall	0.8	0.40	0.69	0.51

We show how our model performs in terms of identifying anomalies—edges that have been injected in the dataset—as we vary the anomaly parameters, \(\pi\). Here \(\mu =0.5\) and \(\rho _{a} = 0.33\)