Identifying abnormal patterns in cellular communication flows

Analyzing communication flows on the network can help to improve the overall quality it provides to its users and allow the operators to detect abnormal patterns and react accordingly. In this paper we consider the analysis of large volumes of cellular communications records. We propose a method that detects abnormal communications events covering call data record volumes, comprising a country-level data set. We detect patterns by calculating a weighted average using a sliding window with a fixed period and correlate the results with actual events happening at that time. We are able to successfully detect several events using a data set provided by a mobile phone operator, and suggest examples of future usage of the outcome such as real time pattern detection and possible visualisation for mobile phone operators.