Title: Evaluation of the queueing network equilibrium based on clustering analysis and self-organising map
Authors: Dimitar Radev, Izabella Lokshina, Svetla Radeva
Addresses: Department of Communication Techniques and Technologies, University of Rousse, 8 Studentska Street, 7017 Rousse, Bulgaria. ' Division of Economics and Business, SUNY Oneonta, 324B Netzer Administration Bldg., Oneonta, NY 13820, USA. ' Department of Computer Aided Engineering, University of Architecture, Civil Engineering and Geodesy, 1 Hr. Smirnensky Blvd. 1046 Sofia, Bulgaria
Abstract: In this paper a new approach to the network local-balance estimation is presented on the base of clustering and neural modelling, in which the cluster centres can replace the steady states of the Markov chain. The equilibrium of queueing networks is a mean for performance evaluation of real communication systems presented as Markov chains. The new approach was developed for local-balance performance evaluation, which was based on calculations of the steady-state probability vector with use of neural modelling and clustering. The experimental results are obtained with use of clustering analysis and self-organising map, and the proposed method is compared with direct techniques on the base of obtained results.
Keywords: mobile communications; mobile networks; Markov reward models; neural networks; queueing networks; clustering; cluster analysis; self-organising maps; local balance estimation; performance evaluation.
World Review of Science, Technology and Sustainable Development, 2007 Vol.4 No.2/3, pp.150 - 160
Published online: 08 May 2007 *Full-text access for editors Access for subscribers Purchase this article Comment on this article