Hybrid Support Vector Machine for Predicting Accuracy of Conflict Flows in Software Defined Networks

Abstract

Software Defined Networking (SDN) is an infrastructure platform for delivering simplified and compliant services with flexible services. These are the means of centralized maintenance and adaptive functions. SDN is affected by various contention flows and causes network performance issues. In this case, we need to provide efficient solutions to handle conflicting flows with better priority and actions. In this paper, we propose a DeepQ Residue method for analyzing normal and conflicting flow scenarios in the load balancing phase. During simulation, an open SDN network is generated using TensorFlow. We use a Hybrid Support Vector machine with an improved decision tree method to predict accuracy and performance. In this case, we analyze threads from 1000 to 100000 in increments of 10000 threads in each iteration. Here, we train a deep belief network with a decision-free feature for environmental simulation. Based on the simulation results, the accuracy of our proposed method reaches 97%, and we compare the results with the results of various existing methods. Our proposed algorithm provides a high-performance SDN application with different conflicting load-balanced flows.