Mining Association Rules For Alarm Data In Businesss Supporting Network Based On Parallel FPGrowth Algorithm

Nowadays,with the rapid development of the communications industry,and intense competition among operators, network maintenance becomes critical for operators. Business support network is used to provide maintenance and support for the operator’s network.In order to betterly carry out the operation and maintenance work, network management is particularly important,with these features of a large-scale network operators, complex structure and equipment varied,which makes the alarm types become very rich.On the other hand,during the network operation and maintenance management, along with the network fails,properly determining the location of network fault,the type of failure and the cause of the problem becomes particularly important in the shortest time,the purpose is to timely repair network faults. However, in the actual network, a failure produces, often followed with multiple alarm events. Moreover, with the continuous improvement of the size and complexity of the network, the type and number of alarms will become more and more.So for network maintenance personnel, only with artificial alarm analysis can not meet with the operation and maintenance work needs. Therefore, this paper proposes to use association rules algorithm of data mining to analyze alarm data of the business support network, the purpose is to pick up valuable information alarms, and use this information to find the root cause of the alarm caused by the fault.The main work of this paper is to build a distributed environment based on Hadoop, while using parallel FPGrowth algorithm to mine association rules for business support network alarm data in this environment. Firstly, preprocessing the alarm data,and use the sliding time window algorithm transform alarm data into alarm transaction data which is suitable for alarm association rule mining; and then using parallel FPGrowth algorithm to mine frequent itemset with the processed data;and finally obtaining association rules from the frequent itemset. as insufficient memory and computing power usually face with bottlenecks during the practical application, so this paper proposes parallel FPGrowth algorithm based on Hadoop platform, combining the advantages of Hadoop platform and MapReduce programming model.