Mining alarm database of telecommunication network for alarm association rules

Some alarm correlation rules were discovered from the alarm databases of telecommunication network management systems by using technology of association rule mining and alarm sequence analysis. The generation and the release information of alarms was recognized and extracted from alarm sequences. The correlation between alarm time and clear time of alarms was used to reorganize alarm data so that better mining result could be obtained. This paper proposes a method to combine the correlated alarms to form an alarm transaction. The method not only solves the problem of asynchrony of correlated alarms in the network, but also discards many redundant data, and reduces the volume of alarm data and the number of rules in the result. The primary experiment shows our method is practical. The discovered rules are useful in locating problems in the network and filtering redundant alarms.