Using dynamic neural networks for server load prediction
Date
2018-06-25
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Lviv Polytechnic National University
Abstract
In this paper, the approach of using neural networks for making time
series predictions of strongly nonlinear data is used. A brief examination of
neural networks usage for time series predictions is given, as well as the
definition and schematics of LSTM blocks. The comparison between the time
delay values and the prediction accuracy is given. It is shown that certain values
of time delay can greatly increase the prediction accuracy.
Description
Keywords
neural networks, long short-term memory, time series prediction
Citation
Pukach P. Using dynamic neural networks for server load prediction / Pavlo Pukach, Volodymyr Hladun // Computational linguistics and intelligent systems, 25-27 June 2018. — Lviv : Lviv Polytechnic National University, 2018. — Vol 2 : Workshop. — P. 157–160. — (Section II. Intelligent Systems).