Today is the era of multimedia applications such as audio or video transmission. Multimedia traffic can be classified according to bandwidth, latency and Quality of Service. For multimedia applications, there is strong need of higher transmission capacity and higher bandwidth, which leads the need of high speed network based architecture at the forefront of several current technology based networks. Due to the unpredictable fluctuations and burstiness of traffic flows within high speed network congestion can occur frequently and TCP will perform poorly in future high speed network. In this paper, we purposed the solution for the problem of frequent congestion occurrence in high speed network and it's model that has learning capability on the basis of round trip time, node's outlink capacity and average queue size and passes a signal to the sender to change its sending rate accordingly. Therefore, congestion occurrence in high speed network can be resolved.