Abstract
Network monitoring is one of the key approaches in network management which can help provide higher network availability; reduction in network operation cost; avoidance of network bottlenecks, and improved efficiency and security. One of the important network performance parameters is time delay across network paths; which is changeable and not a fixed value. Some of these changes are expected and reflect changes in network loading. Other changes are anomalous as they are due to some changes in the network behaviors. We define these changes as Delay Data Exceptions. These Exceptions provide a useful summary of network performance. In this paper, some information about external events were used by Neural Network approach to predict where and what types of the Delay Data Exceptions will be occurred due to these events. The approach is very useful in predicting which part of the network will be affected by a prescheduled unavoidable link interruption. This approach was evaluated by using Delay Data Exceptions generated from two experimental test networks built in the laboratory. The output performance of the Neural Network showed good results