An Intelligent Alarm Management System for Large-Scale Telecommunication Companies

This paper introduces an intelligent system that performs alarm correlation and root cause analysis. The system is designed to operate in largescale heterogeneous networks from telecommunications operators. The proposed architecture includes a rules management module that is based in data mining (to generate the rules) and reinforcement learning (to improve rule selection) algorithms. In this work, we focus on the design and development of the rule generation part and test it using a large real-world dataset containing alarms from a Portuguese telecommunications company. The correlation engine achieved promising results, measured by a compression rate of 70% and assessed in real-time by experienced network administrator staff.