SURVEY ON HETEROGENEOUS NETWORK TRAFFIC ANALYSIS WITH SUPERVISED AND UNSUPERVISED DATA MINING TECHNIQUES?

Network Traffic Analysis (NTA) in heterogeneous networks is one of the emerging research areas receiving substantial attention from both the research community and traffic analyzers. Many tasks in NTA can be naturally cast in a supervised and unsupervised learning model. Many supervised classification models and unsupervised clustering learning models in data mining have been proposed for heterogeneous network. Due to the importance of network traffic analysis in data mining research with the rapid development of new models, To provide a comprehensive review on supervised classification and unsupervised clustering model on heterogeneous type of network in this paper and systematically give a summarization of the state-of-the-art techniques for network traffic analysis. It addresses the problem of network management such as traffic load, quality of service, and trend analysis. This survey covers real time supervised classification and unsupervised clustering algorithms and analyze techniques for heterogeneous networks. It provides taxonomy of the different supervised classification algorithms and unsupervised clustering algorithms and evaluates the various performance metrics that are significantly used for the purpose of comparison. A detailed review is provided covering fuzzy relational clustering algorithm, classification learning algorithms, global voting algorithm and hybrid algorithms. The survey evolve certain open issues, key research challenges for network traffic analysis using supervised classification and unsupervised clustering model in heterogeneous networks, and likely to provide productive research directions.