Statistical Learning and Multiple Linear Regression Model for Network Selection using MIH - Université de Bourgogne Accéder directement au contenu
Communication Dans Un Congrès Année : 2014

Statistical Learning and Multiple Linear Regression Model for Network Selection using MIH

Résumé

A key requirement to provide seamless mobility and guaranteeing Quality of Service in heterogeneous environment is to select the best destination network during handover. In this paper, we propose a new schema for network selection based on Multiple Linear Regression Model (MLRM). A horough investigation, on a huge live data collected from GPRS/UMTS networks led to identify the Key Performance Indicators (KPIs) that play the most important role in the handover process. These KPIs are: Received Signal Code Power (RSCP), received energy per chip (Ec/No) and Available Bandwidth (ABW) of the destination network. To extract a handover model from collected data, we study the correlation among values of identified KPIs parameters, before, during and after handover, thanks to a statistical learning approach, using the predictive analytics software SPSS. For model assessment, Pearson Correlation Coefficient and determination coefficient R-squared (R2) are used. Media Independent Handover (MIH) IEEE 802.21 standard is used in this work to retrieve the lower layer information of available networks and announce the handover needs (handover initiation). The proposed model will help to select the most appropriate network between many existing ones in the vicinity of the mobile node.
Fichier non déposé

Dates et versions

hal-01081434 , version 1 (07-11-2014)

Identifiants

  • HAL Id : hal-01081434 , version 1

Citer

Ahmad Rahil, Nader Mbarek, Mirna Atieh, Olivier Togni, Ali Fouladkar. Statistical Learning and Multiple Linear Regression Model for Network Selection using MIH. The Third International Conference on e-Technologies and Networks for Development (ICeND2014), Apr 2014, Beirut, Lebanon. pp. 195-200 / ISBN: 9978-1-4799-3165-1 ©2014 IEEE. ⟨hal-01081434⟩
123 Consultations
0 Téléchargements

Partager

Gmail Facebook X LinkedIn More