Network Alarm Correlation Analysis Based On Multi-Layer Fuzzy Association Rule Mining

The modern society is an information-oriented society, people’s daily life has beeninseparable from the information technology. Communication network is the basis ofthe information technology, when the communication network fails to work,, whichseriously interferes people’s business life and daily life. So, in order to ensure thenormal and stable operation of network, a strong, efficient fault detection and solutionsmethod is needed, which will timely and accurately detect fault and restore normal andstable operation of network, minimize the interference caused by network failure to theuser.When network fails to work, neighbouring notes will produce lots of alarms. Thekey of network fault diagnosis is to find out the correlation between the alarms, but howto identify this correlation between massive alarm data is one of the difficulties ofcurrent research. Because of the unique advantages of data mining technology on themassive data processing, in recent years, scientists have done lots of research onnetwork fault diagnosis based on association rules of data mining technology, thoughhave achieved a certain amount of research achievement, however, there are still someshortcomings:In one hand, the alarm and fault source has a vague correspondencerelationship, traditional method does not take into account of this ambiguity, but divisesthe alarm into one correspondence relationship between the alarm and faultsource,which ultimately impacts the accurate diagnosis of network failure. In the other hand,communication network has a hierarchical structure, which determines the propagationcharacteristics of the network alarms, so that there is a certain dependency of alarmsbetween layers, but the traditional method rarely consideres this relationship.Considering the above matters and the layer structure characteristics of network,combining fuzzy technology and fuzzy logic with association rule mining techniques,the paper has done the research on network fault correlation analysis method based onfuzzy association rule mining. In the paper, firstly, a network alarm fuzzy algorithm isproposed, secondly, a new multi-layer fuzzy association rule mining algorithm based onthe above algorithm is proposed, which proposes a new fuzzy support computing solutions and fuzzy minimum support setting strategy.Finally, experimental simulationsare done on the proposed algorithms, analyzing the effectiveness, running efficiency,accuracy of the algorithms in-depth. Simulation results tell us that algorithms caneffectively and accurately finish the correlation analysis between the alarms and canprepare alarm accuracy data for final network fault diagnosis.