An adaptive fuzzy traffic estimator of reservation protocols on IP-based networks

Traffic estimation is significant to ensure reliability and the utilization of network resources in the next generation Internet. However it is difficult to estimate network traffic with variable bit rates and bursty data flow. An adaptive approach based on a fuzzy clustering technique for traffic estimation in the resource protocol is proposed. The fuzzy clustering scheme is deployed to estimate traffic flows and predict the data flow in future. All flow patterns are clustered to extract the knowledge about the traffic flows. Instead of adopting a conventional cluster matching scheme for the traffic pattern clusters, a characteristic neural network (CNN) is generated to fuse the obtained clusters in a CNN adaptively to reduce computation. Experiments demonstrate that the proposed approach works well for variable bit rate flows.