Network Alarm Flood Pattern Mining Algorithm Based on Multi-dimensional Association

In the process of network operation, a large number of alerts are generated every day, which reflect the occurrence of some abnormal conditions. Traditional methods depend too much on the knowledge of equipment manufacturers and industry experts, so we need some novel ways to overcome this problem in the network management. The application of data mining technology to alarm pattern analysis has become the focus of current research. Researchers developed many kinds of algorithms fitting different application characteristics. This paper proposes the concept of association matrix pattern mining, which means that before mining the data, we use the multi-dimensional information of the data to construct the association matrices between the items. And we develop a conditional pattern mining algorithm based on the association matrix which aims to find out less but more meaning results. Our experiments validate that with the multi-dimensional information stored in association matrix, the algorithm performs better than traditional pattern mining methods in finding out the detailed alarm pattern from network alarm flood.