The Application and Simulation of NVP Technique in Network Fault Identification
Guyu Hu · Acta Simulata Systematica Sinica · 2003
One of the important means of network fault management is to detect and identify the faults by analyzing the performance data of networks. A solution employing neural network (NN) is proposed in this paper. This solution is developed for the fault identification in the large-scale networks,and it detects and identifies faults through analyzing statistical characteristics of numerical data associated with the network performance. In order to improve the system performance, the software engineering concept of diversity is introduced. NNs are subsequent to training, and are combined by means of a majority voter to form a system. At the same time, the concept of failure correlation coefficient between NNs is defined, and the NNs with negative failure correlation coefficient are selected according to the results of computer simulation, with which the system has achieved 100% generalization to the test sets.