STRUCTURAL OPTIMIZATION OF NEURAL NETWORK FOR QUALITATIVE EVALUATION METHOD OF IT-INFRASTRUCTURE FUNCTIONING

Authors

DOI:

https://doi.org/10.20535/2411-2976.22015.36-43

Keywords:

IT infrastructure, neural networks, the control actions, quality estimation.

Abstract

Background. For today, the unit neural networks are widely used to solve various problems. In this regard, the issue of developing learning algorithm that would be able to optimize the structure of neural networks dynamically is very important. The existence of such a method would allow the researcher to get the structure of the neural network that would be available input data quickly. IT infrastructure allows an organization to deliver IT solutions and services to its employees or customers and is usually internal to an organization and deployed within owned facilities
Objective. In most cases, organizations are able to manage individual elements within their IT infrastructure. But, previously, it is necessary to estimate current quality level of service (QoS) or IT infrastructure functioning generally. In a complex IT infrastructure with mutual influence of its elements it is hard to estimate a quality or its operating.
Method. Taken into a large number of IT infrastructure elements it is important to choose a structure of neural networks automatically. The proposed method makes it possible to control processes inside IT-infrastructure and to form the control actions taking into account the quality of the functioning of IT infrastructure components which makes it advisable to use a control loop
targeted at improving the quality indicators of the performance of the IT-infrastructure.
Results. This paper proposes to use neural networks to evaluate a quality of IT infrastructure functioning. Since the task of determining the structure of the neural network is almost impossible, because it requires a deep analysis of each process taking place in the IT infrastructure, the paper proposes to define the structure of the neural network automatically using structural optimization algorithm of neural network. Series of experiments constructed algorithm that demonstrate the ability to use it in problems of classification of data.
Conclusions. The proposed method makes it possible to control processes inside IT-infrastructure and to form the control actions taking into account the quality of the functioning of IT-infrastructure components which makes it advisable to use a control loop targeted at improving the quality indicators of the performance of the IT-infrastructure. Also the resulting structure of neural
network can be used in a quality estimation of functioning of similar IT infrastructure elements. This will allow the service provider “on the fly” construct and retrain its existing models in a shorter period.
Key words: IT infrastructure; neural networks; the control actions; quality estimation.

References

Scheibenberger, K. Modelling dependencies of IT Infrastructure elements / Scheibenberger, K., Pansa, I. // Business-driven IT Management, 2008. BDIM 2008. 3rd IEEE/IFIP International Workshop on 7-7 April 2008 – p. 112–113

Richardson, P. Quality of service support for wireless Internet service providers / Richardson, P., Sieh, L., Ganz, A. // Computer Communications and Networks, 2001. Proceedings. Tenth International Conference on 2001. – p. 318–323

da Silva, L.F. An IT Infrastructure Patterns Approach to Improve IT Service Management Quality / da Silva, L.F., Brito e Abreu, F. // Quality of Information and Communications Technology (QUATIC), 2010 Seventh

International Conference on the Sept. 29 2010-Oct. 2 2010 – p. 171–176

Toguyeni, A. Quality of service of internet service provider networks: State of the art and new trends / Toguyeni, A., Korbaa, O. // ICTON Mediterranean Winter Conference, ICTON-MW 2007 – 2007 – p. 1–8

Donko, D. Improvement of the process quality in the service provider organization // Telecommunications (BIHTEL), 2014 X International Symposium on – 2014 –p. 1-5

Telenyk S. Qualitative evaluation method of ITinfrastructure elements functioning // Telenyk S., Rolick O., Bukasov M., Dorogiy Y., Halushko D., Pysarenko A. / IEEE International Black Sea Conference on Communications and Networking (BlackSeaCom) / 2014, P 170-174

Hiroshi Ninomiya, An Improved Online quasi-Newton method for robust training and its application to microwave neural network models // Neural Networks (IJCNN), The 2010 International Joint Conference on 18-23 July 2010 – p. 1-8

Hiroshi Ninomiya Parameterized online quasi- Newton training for high-nonlinearity function approximation using multilayer neural networks / Hiroshi Ninomiya // Neural Networks (IJCNN), The 2011

International Joint Conference on July 31 2011-Aug. 5 2011 – p. 2770 – 2777

Qazi, N. Estimation of weights to combine trained neural networks using linear estimation techniques / Qazi, N.// Multitopic Conference (INMIC), 2011 IEEE 14th International – 2001 – p 1-7

Hayashida, T. Structural optimization of neural network for data prediction using dimensional compression and tabu search / Hayashida, T., Nishizaki, I. ; Matsumoto, T. // Computational Intelligence & Applications (IWCIA), 2013 IEEE Sixth International Workshop on 13-13 July 2013 – p. 85–88

Mezard M., and Nadal J.P. Learning in feedforward layered networks: The Tiling algorithm // Journal of Physics. – 1989. – V. A22. – P. 2191 – 2203, Frean M. The Upstart Algorithm: A Method for Constructing and Training Feed- Forward Neural Networks // Tech. Rep. 89/469, Edinburgh

Univ., 1989.

Ash T. Dynamic Node Creation in Back- Propagation Networks // Connection Science. – 1989. – V. 1. Mozer M.C., Smolensky P. Skeletonization: a technique for trimming the fat from a network via relevance assessment // Advances in Neural Information Processing Systems. – 1989. – V. 1. – P. 107 – 115.

Dorogyy Y.Y. Uskorennyiy algoritm obucheniya svertochnyih neyronnyih setey (Accelerated learning algorithm of Convolutional neural networks) /Y.Y.Dorogiy // Visnyk NTUU «KPI». Informatyka, upravlinnya ta obchyslyuval'na tekhnika. – K.: «VEK+»

Hayashida, T. Structural optimization of neural networks and training data selection method for prediction / Hayashida, T., Nishizaki, I. ; Sekizaki, S. ; Nishida, M. // Computational Intelligence and Applications (IWCIA), 2014 IEEE 7th International Workshop on 7-8 Nov. 2014 – p.

–176

Telenyk S.F. Resource allocation and load management considering the assessed quality of the provided service and the use of agent technologies / S.F. Telenyk, A.I. Rolik, A.A. Pokotylo // 23rd International Crimean Conference “Microwave & Telecommunication Technology” (CriMiCo’2013). 9–13 September, Sevastopol, Crimea,

Ukraine. – 2013. – p. 535–536.

Downloads

Issue

Section

Статті