Feature Extraction of Real-World Traces Using K Means Clustering In OppNet

Abstract Natural calamities leave people helpless by arising several situations such as network breakdown, zero communication, intermittent connectivity, dynamic network topology. In such situation an application of dynamic and intermittent routing scheme is essential to make further communication possible during likewise scenarios. An application of TCP/IP becomes futile in mentioned circumstances as it best works for static nodes and pre-defined network topology wherein source and destination nodes are first establishing the communication link with each other. An alternative measure of such hitches is to encounter an application of DTN protocol which possess all characteristics to withstand in such scenarios such as; dynamic network topology, intermittent connectivity, frequent path breaks, store – carry – forward fashion. In this paper we did thorough investigation of forest fire dataset (Uttarakhand) after exploring its implementation in ONE with Epidemic, Prophet, Spray and Wait, HBPR, GAER respectively. An extensive and thorough investigation for real world traces implementation has been done with OppNet routing protocols against mobility models namely; Shortest path map – based, Random Direction, Random Walk, Random Waypoint, Cluster Movement respectively for network performance metrics namely packet delivery ratio, packet overhead ratio and average latency ratio respectively with the application of K means clustering machine learning algorithm. With the help of this analysis, we explore the real-world traces characteristics and study the areas on which network performance can be improved.