The last years have seen a continuous increase in the use of mobile communications. To cope with the growing traffic, recently deployed technologies have deepened the adoption of small cells (low powered base stations) to serve areas with high demand or coverage issues, where macrocells can be both unsuccessful or inefficient. Also, new cellular and non-cellular technologies (e.g. WiFi) coexist with legacy ones, including also multiple deployment schemes (macrocell, small cells), in what is known as heterogeneous networks (HetNets).
Due to the huge complexity of HetNets, their operation, administration and management (OAM) became increasingly difficult. To overcome this, the NGMN Alliance and the 3GPP defined the Self-Organizing Network (SON) paradigm, aiming to automate the OAM procedures to reduce their costs and increase the resulting performance. One key focus of SON is the self-healing of the network, covering the automatic detection of problems, the diagnosis of their causes, their compensation and their recovery.
Until recently, SON mechanisms have been solely based on the analysis of alarms and performance indicators. However, on the one hand, this approach has become very limited given the complexity of the scenarios, and particularly in indoor cellular environments. Here, the deployment of small cells, their coexistence with multiple telecommunications systems and the nature of those environments (in terms of propagation, coverage overlapping, fast demand changes and users' mobility) introduce many challenges for classic SON.
On the other hand, modern user equipment (e.g. smartphones), equipped with powerful processors, sensors and applications, generate a huge amount of context information. Context refers to those variables not directly associated with the telecommunication service, but with the terminals and their environment. This includes the user's position, applications, social data, etc.
