Cellular networks have been increasing in size and complexity constantly since the earliest generations. This growing complexity makes it harder for network operators to manage and improve the efficiency of the network while maximizing the quality of experience (QoE) of its users. As a way to ease the management of such complex networks, self-healing and automatic network-optimization methods have been developed over the years. Implementation of this methods made networks capable of troubleshooting problems previously identified by network experts, reducing the work effort required to maintain a high QoE. To automatically identify these network problems, unsupervised classification techniques have been put to use, since the amount of labelled data required for supervised techniques is not always available or complete. This paper proposes a method based on multi-resolution analysis and clustering for the detection and identification of anomalies in cellular networks through different Key-Performance Indicators (KPIs).