A Closed-Loop Automatic Data-Mining Approach for Preventive Network Monitoring

作者：Caravela Ivan; Arsenio Artur^*; Borges Nuno

来源：Journal of Network and Systems Management, 2016, 24(4): 974-1003.

DOI：10.1007/s10922-015-9356-6

• Data mining
• Machine learning
• Target event
• Knowledge base
• Association rules
• Training
• Sequence classification
• EventSets
• Markov chains

As telecommunication networks grow in size and complexity, monitoring systems need to scale up accordingly. Alarm data generated in a large network are often highly correlated. These correlations can be explored to simplify the process of network fault management, by reducing the number of alarms presented to the network-monitoring operator. This makes it easier to react to network failures. But in some scenarios, it is highly desired to prevent the occurrence of these failures by predicting the occurrence of alarms before hand. This work investigates the usage of data mining methods to generate knowledge from historical alarm data, and using such knowledge to train a machine learning system, in order to predict the occurrence of the most relevant alarms in the network. The learning system was designed to be retrained periodically in order to keep an updated knowledge base.

• 出版日期2016-10