Mobile Alarm Correlation System Based On Data Mining

With the rapid expansion of the mobile network along with the increase of usersand businesses, the amount of equipment alarms are growing sharply on net-watchscreen due to component maintenance, system optimization, and device faults. It isthe main job of monitoring personnel to accurately find the key information from alot of alarms. Without the aid of analysis tools, they fail to locate the fault and solveit quickly and accurately according to their experience. To solve the problem, thispaper analyzes alarm data to find their relationship, and builds the models of alarmcorrelation. Then the models are applied to net watcher, helping the monitoringpersonnel with the accurate the fault position as soon as a alarm appears.First, the data mining technology is applied to alarm correlation. In mining,there are three components in mining alarm database: data selection, datapreprocessing, and transformation. After the preprocessing data areobtained, thePF-growth algorithm is employed to mine the alarm data. In order to obtain thefrequent itemsets to built the model of the alarm correlation, the five-months alarmdata are used in Hunan mobile company. By mining the preprocessed data, twocorrelation models, primary-subordinate-relationship and deriving relationship, areobtained.In order to realize the alarm correlation on the net watcher, we decided to addthe correlation module, which can correlate alarms and send it to the net wathermonitor termianal. Function realization is divided into three parts. First: through thesearch engine module of the alarm correlation, system accepts the original alarm andconfigures data to build the relation or release the relation.Second: through theLocation information parser, system parses the alarm to obtain the availabilityinformation, which is needed when alarm correlated.Thrid: through the standardclient, system display the alarm correlation.All the fault can be displayed in the net watcher simply and perfectly. After thereal-time alarms are associated, monitoring personnel can judge and position thefaults in a short time.At the same time, the system can abstract and refine the alarmdisplaying on the net watcher.