Implementation of Loss Discrimination Algorithm Using Machine Learning Approach for Wireless TCP Enhancement

Abstract

Abstract

In wired network packet losses are due to network congestion whereas in wireless network, it can be due to variable bandwidth, network topology, host mobility, network error etc. Transmission Control Protocol (TCP) is used to handle wired and wireless network. This conventional TCP consider any loss as congestion loss whatever may be the reason behind loss and to control the loss it simply reduces the congestion window size (Cw) which leads to deteriorate the network performance due to low utilization of bandwidth. Thus to improve the performance of wireless TCP, we are introducing loss discrimination algorithm (LDA) based on machine learning (MC) which will classify the losses in to two categories as congestion loss and random error (viz. high bit error rate, link problem etc.) packet loss. Further based on the type of loss the corrective action is taken. This avoids the unnecessary reducing the congestion window size in turn improves the network performance. Proposed system (ML-LDA) is an extension of TCP Reno which was developed to improve the performance of conventional TCP and we are getting better performance than TCP Reno.