On association rules mining in the context of wireless networks

Abstract

To assess the performance of a cellular wireless network such as GSM, UMTS and LTE etc., performance counters are logged and maintained by the network management system (NMS). Due to the complexity of a network, the number of these performance counters is typically very large, and the analysis of these data is very difficult. In a typical situation, only a few number of key performance indicators (KPIs) are used to represent the performance status of the network. As a result, the vast amount of data contained in the NMS database is not fully utilized. If the network operator can exploit such data more fully, more insights regarding the network behaviours can be obtained and extracted. These insights are also very helpful in diagnosing network problems. This paper proposes applying what is known as the market basket analysis to analyse the network KPIs. The bridging of these two seemingly unrelated domains is done via the use of association rules mining. An example is presented to illustrate the usefulness of this approach.