Proactive network maintenance using machine learning

We describe a new approach to preactively maintain a massively interconnected communications networks over time. We have applied this approach to the detection and prediction of chronic transmission faults in AT&T´s digital communications network. A windowing technique was applied to large volumes of diagnostic data and these data were analyzed by machine learning methods. A set of conditions has been found that is highly predictive of chronic circuit problems, that is, problems that are likely to continue in the immediate future without diagnosis and repair. In addition, a few conditions have been found that are predictive of problems that affect multiple circuits. Such analyses over the complete network can be helpful in proactively maintaining the network and in spotting trends for circuit problems. Proactive maintenance of the network can help in greatly improving the quality and reliability of a network by identifying potentially serious problems before they degrade