The Study Of Optical Transmission Network Fault Diagnosis Based On Data Mining

With the development of telecommunication network, the scale of which is more and more huge and the structure is more and more complex. There are also diverse equipments in the telecommunication networks, leading to informative alarms and complicated alarms relationship.Therefore, fault diagnosis has become increasingly difficult, and the cost of diagnosis is increasing as well. In order to solve this problem,we need to analyze the correlation between the alarms, mining association rules between the alarms. Finding the source of the fault alarms according to the rules, can not only reduce the work of the management, but also minimize the losses.The study of association rules is one of the main branches of data mining. Apriori algorithm is a classical algorithm of association rule mining. This paper mainly studies the alarm correlation approaches, Apriori association rules algorithm and parallel programming model. We improve Apriori algorithm and implement it based on OpenMP in a parallel manner to achieve enhanced performance, the proposed Apriori algorithm is called weighted association rules algorithm.The contribution I made independently in this paper are mainly in the following four parts. Firstly, to solve the problem of repeatedly scan the database when generating frequent itemsets in Apriori algorithm, we used more spaces to reduce the number of scanning.. Meanwhile, we put the frequency of occurrence in the database as the weight of the items, so as to enhance the importance of the high frequency items. Secondly, this paper adopted the ’and’ operation between the matrixes, without generating candidate itemsets, avoiding large number of frequent itemsets, which greatly improved the efficiency of mining. Thirdly, Because the alarm data and the data matrixes cannot stored in the memory all the time, we introduced the OpenMP multi-thread shared memory parallel programming model to implement the algorithm in a parallel manner. By the parallel optimization, the time cost of the numerical calculation is reduced and the performance of the algorithm has been improved. Fourthly, this paper designed the optical transmission network diagnosis system model, which are also applied in Sichuan Unicom transmission network system. Our proposed model has been verified. Finally, this paper summarized our work and the work in the future.