Оценка точности методов краткосрочного прогнозирования нагрузки компьютерной сети

Abstract

Nowadays, the nature of the load of the users on computer networks is unpredictable, and the degree of such unpredictability tends to increase. There are also conditions for the constant changes in the architecture of the computer networks due to the increase in the number of the different wireless networks compared to the cable infrastructure in whole. Therefore, the category of the methods of the short-term forecastination seems to be the most suitable and promising methods for predicting the behavior of the network traffic, since this group of methods allows you to do a quickly respond to different ongoing changes in the architecture of the network and to an increase or decrease in user’s load on a computer network. The article provides a review and a brief description of some methods of short-term forecasting. The estimation of the accuracy of the prediction of the considered methods is also given in this paper. This estimation is based on a time series (data array), whose elements were the number of packets in the network per second of time. The data array was obtained as a result of processing a network traffic dump, which collected using the Wireshark computer network analyzer program.