Modern telecommunication systems produce large amounts of alarm messages, and alarm management is vital for telecommunication systems’ high-quality performance. Building functional networks by observing the pair similarity between time series is a useful way to filter and reduce alarm messages. Because of the coexistence of positive and negative correlations among telecommunication devices, most of the similarity measures have troubles in computing the complex correlations. In this paper, we propose an index of measuring how much two-alarm series deviate from the uncorrelated situation to detect the correlation of both sides. Synthetic sequences verify our method. Furthermore, we apply our method to analyze telecommunication devices’ alarm correlation in a province of China. Our index of pair similarities is capable of measuring other discrete event data.
Similarity Analysis of Alarm Sequences by a Shuffling Method