METHOD FOR DETECTING ANOMALIES IN MOBILE TELECOMMUNICATION NETWORKS

A method for detecting anomalies in a mobile network, wherein a time series data input (1) based on performance metrics is obtained for a plurality of network elements. A similarity metric is defined by calculating the squared euclidean distance between network elements to detect (2) clusters of network elements that have the same value of the similarity metric by applying the Louvain algorithm. The clusters with anomalies are determined if the similarity metric differs from the average value in a number N of times the standard deviation. The method checks whether any open ticket/alarm is associated with any network element of the clusters. This improves customer experience without increasing the cost for network operators as each anomaly and its root cause can be identified. Also, continuous learning allows the operator to deploy automatic anomaly detection in a flexible way in their networks.