Mobile data traffic modeling: Revealing temporal facets

Abstract

Rapport de recherche n°8613 version 5 version initiale October 2014 version révisée Juin 2015 32 pages Résumé : Comprendre la demande de trac de données mobiles est essentielle pour l'évaluation des stratégies portant sur le problème de l'utilisation de bande passante élevée et l'évolutivité des ressources du réseau, apporté par l'ère "pervasive". Dans cet article, nous eectuons la première modélisation détaillée de l'utilisation du trac mobile des smartphones dans un scénario métropolitain. Nous utilisons un ensemble de données à grande échelle recueillis au coeur d'un des majeurs réseaux 3G de la capitale du Mexique. Nous analysons d'abord le comportement individuel routinier et nous avons observé des modèle d'utilisation identiques pour les diérents jours. Cela nous motive à choisir un jour pour étudier le mode d'utilisation des abonnés (c'est à dire, "quand" et "combien" de trac est généré) en détail. Nous classons ensuite les abonnés en quatre prols distincts en fonction de leur mode d'utilisation. Nous modélisons enn le mode d'utilisation de * This work was supported by the EU FP7 ERANET program under grant CHIST-ERA-2012 MACACO.

Tactful Networking: Humans in the Communication Loop

“…Browsing (visited websites) behavior has also been applied in user profiling according to their traffic demand [144]. Still, other works have categorized the actual mobile traffic usage [145], [146]. Among those, [146] provides a profiling of individual users' behavior -rather than a network-wide oneand a precise temporal network usage modeling, i.e., in terms of volume as well the frequency of traffic demand -rather than only considering total volume of traffic or a normalized volume.…”

“…Still, other works have categorized the actual mobile traffic usage [145], [146]. Among those, [146] provides a profiling of individual users' behavior -rather than a network-wide oneand a precise temporal network usage modeling, i.e., in terms of volume as well the frequency of traffic demand -rather than only considering total volume of traffic or a normalized volume. Among the outcomes, authors show: (i) the high daywise similarity on sessions number, traffic volume, and interarrival time traffic parameters; (ii) such parameters from the same hours on different days present less variability than the parameters within the same day on different hours; (iii) the high correlation between upload/download traffic volume; (iv) peak and non-peak hours can be easily identified when it comes to users' traffic demands; (v) similar sessions number and duration occur when users are grouped by age range, irrespective of the users' gender; (vi) male participation raises as the user age increases, while the female participation decreases with the age increase.…”

“…More recently, literature works have investigated network usage concerning other users' behavior. In [146], [147], calls/SMS patterns or traffic demand of users and their sociodemographic factors (e.g., age and gender) are jointly investigated. The relation between content consumption and mobility properties is considered in studies that focus on application interests [148], data traffic dynamics [149] and service usages [150].…”

ECMCRR-MPDNL for Cellular Network Traffic Prediction With Big Data

“…But the method failed to ensure higher prediction accuracy. A measurement-driven model was developed in [17] for mobile data traffic prediction using big data collected from a crowded metropolitan area. The traffic parameters spatial correlation was not analyzed to further enhance traffic prediction.…”

Snel: SQL Native Execution for LLVM

“…The social graph constructed from the communications between mobile phone users is rich in information that was leveraged to make inferences such as dynamic communities [23], inferences of age, gender and other socio-demographic attributes [24,25,26,27,5]. In particular it provides an effective tool for predicting customer churn [28,29,30], along with a variety of other predictions [31,32,33,34,35].…”