Prediction of TCP Congestion by Using Machine Learning

Congestion, inside the context of networks, refer to a network state where a node carries lot of data that may deteriorate network service quality, which results in blocking of new connections. Congestion occurs when bandwidth is insufficient and network data traffic exceeds its limit. Even though many TCP variants exists congestion control is still an issue. Here there is new method of measuring network congestion passively which gives an accurate result of congestion by using machine learning algorithm. The prediction of TCP congestion window help us to avoid congestion. Here we are utilizlizing dataset of TCP variants and from iperf tool which captured packets when transfer in between of source and destination in real time environment. On this dataset we apply Random Forest Algorithm on training dataset and on testing dataset. The correctness of result is shown in term of accuracy, Mean Square Error (MSE). In this experiment we found 84% accuracy and 8.00% Mean Square Error (MSE).