A neural network apporach to alarm correlation

At the heart of network fault management is the alarm correlation. The paper introduces the basic theory of network alarm correlation. It then analyzes the limitations of some other artificial intelligent correlation methods. An approach based on neural network techniques is presented to solve alarm correlation. Back propagation arithmetic is used to regulate the weight of neural network. As the example of a link fails in the real network, the correlation approach is simulated in the environment of Matlab. The results show that this approach can correlate correctly with noise within a limited range. It has advantages in dealing with uncertain information and in restraining noises, and has a certain practical value.