Akademik Veri Yönetim Sistemi
ADVANCES IN ELECTRICAL AND COMPUTER ENGINEERING, cilt.23, sa.3, ss.57-66, 2023 (SCI-Expanded)
the increase in the number of devices using local and global networks has made it very challenging to manage network traffic, especially during epidemics and natural disasters. Traffic must be analyzed and routed efficiently while managing the network in these situations. The proposed approach focuses on effective routing by detecting elephant flows. In this study, the Deep Learning method has been utilized for elephant flow detection. In flow detection, 11 different features are extracted for each flow, and elephant flows are accurately detected. Additionally, the Grid Search method determines the parameters that yield the best results in classifying elephant and mice flows. As a result, elephant flows that have been classified are routed using the Discrete-Particle Optimization method, whereas mice flows are routed using traditional Round Robin or Random methods. The experimental results show that the proposed approach achieves a high level of accuracy in detecting elephant flows and routing them effectively while also maintaining the overall network performance.