Нейро-нечеткое прогнозирование длительности ожидания подтверждений в телекоммуникационной сети

Abstract

Article is devoted to the development of a system for predicting of confirmations waiting time in the telecommunication network. The input of the values of the waiting time served on three previous acknowledgment packet sent data source. The system is based on the use of four-fuzzy neural network functioning on the basis of eight fuzzy rules. The functional diagram of the system is submitted. To configure the system formed training data obtained by measuring the waiting time on the packets acknowledgments predetermined data stream in real telecommunication network. Numerous simulation experiments performed in the software environment Matlab, showed an acceptable waiting time prediction accuracy using the synthesized evidence of neuro-fuzzy systems. The proposed system can be used to estimate the congestion telecommunications network, as well as to control the intensity data and sending retransmissions.