Abstract
With the widespread use of over-the-top (OTT) media, such as YouTube and Netflix, network markets are changing and innovating rapidly, making it essential for network providers to quickly and efficiently analyze OTT traffic with respect to pricing plans and infrastructure investments. This study proposes a time-aware deep-learning method of analyzing OTT traffic to classify users for this purpose. With traditional deep learning, classification accuracy can be improved over conventional methods, but it takes a considerable amount of time. Therefore, we propose a novel framework to better exploit accuracy, which is the strength of deep learning, while dramatically reducing classification time. This framework uses a two-step classification process. Because only ambiguous data need to be subjected to deep-learning classification, vast numbers of unambiguous data can be filtered out. This reduces the workload and ensures higher accuracy. The resultant method provides a simple method for customizing pricing plans and load balancing by classifying OTT users more accurately.
Highlights
With the advancements of smart devices and the rapid development of wired and wireless networks, our modes of entertainment and venues for information are changing rapidly
For classification based on k-nearest neighbor (KNN), decision tree, support-vector machine (SVM), naïve Bayes, and repeated incremental pruning to produce error reduction (RIPPER), we employed
For the multilayer perceptron (MLP), ActivationELU was used as the activation function of the hidden and output layers, ADAM was used as the optimizer of the loss function, and AdaDelta was used as the bias updater
Summary
With the advancements of smart devices and the rapid development of wired and wireless networks, our modes of entertainment and venues for information are changing rapidly. Sci. 2020, 10, 8476 bandwidth overloads by limiting the amount of throughput (i.e., throttling) based on the OTT service in demand This process is most often reactionary, inevitably creating surges of consumer complaints. We analyze network consumption patterns based on consumers’ OTT-usage patterns, confirming that a combination of machine- and deep-learning capabilities can achieve the highest accuracy. By mitigating the time and resource requirements of deep learning, we provide a novel MetaCost-based framework related to OTT user analysis that can reduce the time required for analysis while exploiting the technology’s high accuracy This framework drastically reduces the analysis workload, making the process very efficient and timely so that ISPs can achieve instantaneous status and influence over OTT service demands.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
