Distributed Association Rule Mining In Telecommunication Alarm Correlation Analysis

Fault diagnosis and localization is the vital core of the network fault management. When the faults take place in the networks, it is necessary to find the locations and the causations of the faults accurately as soon as possible in order to get rid of the faults and recover the networks'function in time. The alarm correlation analysis is an important approach of fault diagnosis, it plays a crucial role in network fault management. Data mining provides a new approach of the knowledge updating during the alarm correlation analyzing. The modern telecommunication networks and its network management is a typical distributed system. The realization of management fuction is done by different management components which in different levels to cooprate with each other. With the background of the project, the alarm correlation in telecommunication networks based on data mining, that is supported by National Natural Science Foundation of China, this thesis focused on distributed alarm association rules mining in telecommunication networks, including the alarm's pretreatment, algorithem of association rules distributed mining and the simulation and validation of the system of distributed alarm association rules mining in telecommunication networks.The problem of alarm's synchronization was settled by setting time window and the slip length. The attributes of the alarm that reflect the faults were picked out to form an item of an alarm transaction and the redundant alarms were got rid of by alarm compressing. The alarm's pretreatment was carried out to transform the alarm database into alarm transaction database which make ready the data for distributed alarm association rules mining.Aimed at the telecommunication networks'characters that are having a lot of alarms that can burst out randomly and the alarm databases are generally located distributedly, the shared pattern distributed structure was introduced, so the global alarm associations are mined coordinated with local stations and the global station. Based on the existent algorithms of mining association rules, a new algorithm named EDMA which suits for distributed alarm association rules mining was proposed. In local stations the algorithm runs the application based on the ameliorated algorithm CMatrix, which was based on a compressed association matrix. Algorithm EDMA is unnecessary to traverse the database repeatedly. By numbering the global frequent itemsets that are generated during each iteration orderly, the algorithm mapps every itemsets into a number-couple correspondingly. By the mapping, EDMA uses the number-couple instead of the itemsets themselves to transmit between local stations and the global station, in this way, the global station can not only use hash table to match and store the candidate itemsets, but also compress the context of itemsets when transmitted. The performance test of the algorithm indicates that EDMA has high running efficiency, low transmission cost and strong retractility.Some technique difficulties when constructing the high performance mining server were settled by adopting IOCP mechanism to realize the communication control module and by applying memory pool to I/O data buffer management and by customizing communication protocol when communicating between local station and the global station. In the end, a integrated system of distributed alarm association rules mining in telecommunication networks was constructed. The simulation and validation of the system indicates that it is able to mining the useful alarm association rules from the alarm database rapidly and effectively.