Real-time application clustering in wide area networks

Takyi, Kate; Bagga, Amandeep

doi:10.1016/j.compeleceng.2020.106691

Cited by 3 publications

(1 citation statement)

References 9 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Clustering algorithms are usually applied to solve problems related to data mining [36]. Clustering is a technique used for grouping data, categorized into unsupervised and semi-supervised [37], with the main goal of classifying a set of data objects into several groups named clusters. The objects of a cluster must have high similarity to each other and must be dissimilar to the objects of other clusters [38].…”

Section: Order Selectionmentioning

confidence: 99%

Packet Loss Characterization Using Cross Layer Information and HMM for Wi-Fi Networks

Silva

Pedroso

2022

Sensors

View full text Add to dashboard Cite

Packet loss is a major problem for wireless networks and has significant effects on the perceived quality of many internet services. Packet loss models are used to understand the behavior of packet losses caused by several reasons, e.g., interferences, coexistence, fading, collisions, and insufficient/excessive memory buffers. Among these, the Gilbert-Elliot (GE) model, based on a two-state Markov chain, is the most used model in communication networks. However, research has proven that the GE model is inadequate to represent the real behavior of packet losses in Wi-Fi networks. In this last category, variables of a single network layer are used, usually the physical one. In this article, we propose a new packet loss model for Wi-Fi that simultaneously considers the temporal behavior of losses and the variables that describe the state of the network. In addition, the model uses two important variables, the signal-to-noise ratio and the network occupation, which none of the packet loss models available for Wi-Fi networks simultaneously take into account. The proposed model uses the well-known Hidden Markov Model (HMM), which facilitates training and forecasting. At each state of HMM, the burst-length of losses is characterized using probability distributions. The model was evaluated by comparing computer simulation and real data samples for validation, and using the log-log complementary distribution of burst-length. We compared the proposed model with competing models through the analysis of mean square error (MSE) using a validation sample collected from a real network. Results demonstrated that the proposed model outperforms the currently available models for packet loss in Wi-Fi networks.

show abstract

Section: Order Selectionmentioning

confidence: 99%