Latent Clustering Models for Outlier Identification in Telecom Data

Figure 3

Anomaly detection with GPLSA from traffic data set presented in Section 5. Plots are restricted to two days in (a), (b), and (c). Red and orange points are related to the lowest likelihood values obtained, with an average of 2 red points and 8 orange points each day.

(a) Values as a function of dates, with clusters identified

(b) Values as a function of dates, with classes identified

(c) Log-likelihood of values of the set as a function of dates

(d) Probability to be in a cluster knowing date class