EANN 2012: exploratory analysis of mobile phone traffic patterns using 1-dimensional SOM, clustering and anomaly detection

• Published in: Evolving systems 2013-12, Vol.4 (4), p.251-265
• Main Authors: Kumpulainen, Pekka, Hätönen, Kimmo
• Format: Article
• Language: English
• Subjects: Artificial Intelligence; Complex Systems; Complexity; Engineering; Original Paper
• ISSN: 1868-6478; 1868-6486
• DOI: 10.1007/s12530-013-9091-8

Mobile network is a continuously evolving system. Recognizing the behaviour of the subscribers helps the operator in managing and developing the services they offer. Valuable information for those purposes can be revealed by analyzing the daily traffic patterns which result from the usage habits of the subscribers. In this paper we show how exploratory data analysis can be used to discover valuable information from the daily patterns. We present generic tools and methods that are especially suitable for data that consists of patterns. The proposed procedure consists of four parts: preprocessing, day of the week analysis, network element analysis and anomaly detection. The methods are based on clustering and one dimensional self organizing map. One dimensional map enables compact visualization that is especially suitable for data where the variables are not independent but form a pattern. We found out that the day of the week is the main explanation for the traffic patterns on weekends. On weekdays the traffic patterns are mostly specific to groups of networks elements, not the day of the week. The results on the mobile network traffic are presented on a general level. Detailed analysis of the root causes behind the findings requires knowledge about the configuration, topology and geography of the network, which are available only for the operator.