Optimizing next-generation mobile networks using frequent sequential pattern mining

Abstract

With the increasing number of mobile users and the high volume of traffic that exists in modern mobile networks, it is evident that new methods for uncovering the patterns that exist within the usage data need to be developed. We present an algorithm called Prefix-based Mobility Mining (PRIMO) that combines frequent sequential pattern mining and association rules to build a system that can be used to improve traffic handling and predict the future actions of users in a network. We also introduce an application called AndroidMiner that explores the applicability of pattern mining to the mobile computing environment. When a mobile user interacts with a network they are often in contact with multiple cell towers. Employing the sequential nature with which the identifiers of these towers can be collected, we can effectively generate frequent sequences that represent the towers a user or group of users are most frequently interacting with. By representing the interactions of mobile users with networks as frequent sequences, we can effectively develop an understanding of traffic patterns that exist within a mobile network. An understanding of the trends of mobile networks can easily be applied to load balancing or improving location-based services.