Cellular Traffic Prediction and Classification: A Comparative Evaluation of LSTM and ARIMA

Azari, Amin; Papapetrou, Panagiotis; Denic, Stojan; Peters, Gunnar

doi:10.1007/978-3-030-33778-0_11

Cited by 64 publications

(42 citation statements)

References 34 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Publicly available data set: We also evaluate the metalearning scheme on four data sets created from the Wireshark traces collected and analyzed in [19], available on GitHub. The data set contains cellular uplink and downlink traffic, of which we only considered the downlink traffic.…”

Section: Numerical Resultsmentioning

confidence: 99%

“…In general, each data set spans over tens of thousands of prediction intervals. In addition we evaluated the predictors on four publicly available data sets reported in [19]. In the appendix, Table IV provides a summary of the data sets.…”

Section: B Evaluation Methodologymentioning

confidence: 99%

“…Compared to the linear models, DNNs provide significant improvements in prediction accuracy, as they contain numerous levels of non-linearities depending on the number of hidden layers, which allow them to efficiently represent highly nonlinear patterns and quickly-varying functional abstractions [18]. The experiments in [19] show superior performance of the prediction tool based on long short-term memory (LSTM) neural networks over the ARIMA models in cellular traffic classification and prediction. The combination of linear and nonlinear models is referred to as a hybrid model, which was found to give good results in the prediction and analysis of network traffic [20].…”

Section: Related Workmentioning

confidence: 99%

“…Originated in econometric analysis, this technique is not very efficient for short term prediction. An extensive review of previous work on network traffic prediction is given in [19], and shows that despite a variety of techniques, the rapidly changing nature of data traffic and its non-stationarity make it difficult and computationally expensive to train a one-size-fit-all predictor.…”

Section: Related Workmentioning

confidence: 99%

See 3 more Smart Citations

A Meta-Learning Scheme for Adaptive Short-Term Network Traffic Prediction

Moayyedi

Dán

et al. 2020

IEEE J. Select. Areas Commun.

View full text Add to dashboard Cite

Network traffic prediction is a fundamental prerequisite for dynamic resource provisioning in wireline and wireless networks, but is known to be challenging due to non-stationarity and due to its burstiness and self-similar nature. The prediction of network traffic at the user level is particularly challenging, because the traffic characteristics emerge from a complex interaction of user level and application protocol behavior. In this work we address the problem of predicting the network traffic at the user level over a short horizon, motivated by its applications in cellular scheduling. Motivated by recent works on robust adversarial learning, we treat the prediction problem for non-stationary traffic in an adversarial context, and propose a meta-learning scheme that consists of a set of predictors, each optimized to predict a particular kind of traffic, and of a master policy that is trained for choosing the best fit predictor dynamically based on recent prediction performance, using deep reinforcement learning. We evaluate the proposed meta-learning scheme on a variety of traffic traces consisting of video and nonvideo traffic. Our results show that it consistently outperforms state-of-the-art predictors, and can adapt to before unseen traffic without the need for retraining the individual predictors.

show abstract

Section: Numerical Resultsmentioning

confidence: 99%

Section: B Evaluation Methodologymentioning

confidence: 99%

Section: Related Workmentioning

confidence: 99%

Section: Related Workmentioning

confidence: 99%

See 2 more Smart Citations

A Meta-Learning Scheme for Adaptive Short-Term Network Traffic Prediction

Moayyedi

Dán

et al. 2020

IEEE J. Select. Areas Commun.

View full text Add to dashboard Cite

show abstract

“…On the contrary, LSTM is ready to process multivariate time series. The authors of [35] and [36] revealed that, the higher the number of input variables, the better the traffic predictions of LSTM when compared to ARIMA.…”

Section: B Time Series For Network Traffic Forecastmentioning

confidence: 99%

Predictive CDN Selection for Video Delivery Based on LSTM Network Performance Forecasts and Cost-Effective Trade-Offs

Viola

Martín

Morgade

et al. 2021

IEEE Trans. on Broadcast.

View full text Add to dashboard Cite

Owing to increasing consumption of video streams and demand for higher quality content and more advanced displays, future telecommunication networks are expected to outperform current networks in terms of key performance indicators (KPIs). Currently, content delivery networks (CDNs) are used to enhance media availability and delivery performance across the Internet in a cost-effective manner. The proliferation of CDN vendors and business models allows the content provider (CP) to use multiple CDN providers simultaneously. However, extreme concurrency dynamics can affect CDN capacity, causing performance degradation and outages, while overestimated demand affects costs. 5G standardization communities envision advanced network functions executing video analytics to enhance or boost media services. Network accelerators are required to enforce CDN resilience and efficient utilization of CDN assets. In this regard, this study investigates a cost-effective service to dynamically select the CDN for each session and video segment at the Media Server, without any modification to the video streaming pipeline being required. This service performs time series forecasts by employing a Long Short-Term Memory (LSTM) network to process real time measurements coming from connected video players. This service also ensures reliable and cost-effective content delivery through proactive selection of the CDN that fits with performance and business constraints. To this end, the proposed service predicts the number of players that can be served by each CDN at each time; then, it switches the required players between CDNs to keep the (Quality of Service) QoS rates or to reduce the CP's operational expenditure (OPEX). The proposed solution is evaluated by a real server, CDNs, and players and delivering dynamic adaptive streaming over HTTP (MPEG-DASH), where clients are notified to switch to another CDN through a standard MPEG-DASH media presentation description (MPD) update mechanism.

show abstract

Active Learning Training Strategy for Predicting O Adsorption Free Energy on Perovskite Catalysts using Inexpensive Catalyst Features

2021

View full text Add to dashboard Cite

Machine learning (ML) based energy prediction models are among the most effective descriptor-based catalyst screening tools for heterogeneous reaction systems. However, their implementations are limited due to expensive data labelling, ab initio feature evaluation and lack of universal catalyst features, that is, beyond d-band theory. Herein, we propose an inexpensive geometric feature for application on systems beyond d-band theory, for example perovskites comprising of s-, p-, d-and f-block elements. We outline a workflow that inputs these features into an active learning algorithm that enables effective data labelling, whilst improving prediction accuracies of existing models. We then use batch sampling to define termination criteria and to implement time-series error forecasting for further reducing the number of expensive data labelling for training. We implement this workflow to train ML models for predicting oxygen adsorption free energy on perovskites and achieve similar, if not better, prediction accuracies as obtained from ab initio features.

show abstract

Cellular Traffic Prediction and Classification: A Comparative Evaluation of LSTM and ARIMA

Cited by 64 publications

References 34 publications

A Meta-Learning Scheme for Adaptive Short-Term Network Traffic Prediction

A Meta-Learning Scheme for Adaptive Short-Term Network Traffic Prediction

Predictive CDN Selection for Video Delivery Based on LSTM Network Performance Forecasts and Cost-Effective Trade-Offs

Active Learning Training Strategy for Predicting O Adsorption Free Energy on Perovskite Catalysts using Inexpensive Catalyst Features

Contact Info

Product

Resources

About