Exploring TCP Congestion Control: A Comprehensive Analysis of Cubic, RENO, and VENO Algorithms Using Hybrid Methodology and Machine Learning

TCP congestion control is essential for maintaining network stability and efficient data transfer in the context of contemporary networking. Three well-known algorithms: Cubic, RENO, and VENO were used in this study to analyze TCP congestion in detail with a specific focus on validating the superiority of the Cubic Algorithm. This work introduces a novel hybrid methodology combining machine learning techniques and human intervention. Congestion pattern analysis was done using 13 CNN models, such as ResNet50V2, ResNet152, and Xception. ResNet50V2 led in validation accuracy at 0.566666663 and Xception in training at 0.459259272. For RNN, GRU had the highest train and test accuracies of 37.21% and 32.47%, respectively, and detected TCP and various other protocols within packet files with a train and test accuracy of 99.29% and 99.51%, respectively. The study concludes by noting the superiority of human intervention over machine learning techniques for TCP congestion control analysis.