Anomaly detection in mobile phone data - Exploratory analysis using Self-Organizing Maps

Communications traffic on wireless networks generates large amounts of metadata on a continuous basis across the various servers involved in the communication session. The networks are engineered for high reliability and hence, the data from these networks is predominantly normal with a small proportion being anomalous. From an operations perspective however, it is important to detect these anomalies when they occur to correct any vulnerabilities in the network. The objective of our work is anomaly detection in communication networks to improve network performance and reliability. In this paper we explore the use of neural network based Kohonen Self Organizing Maps (SOM) applied to Per Call Measurement Data (PCMD) records from a 4G network for data analysis and anomaly detection.