Urban traffic flows forecasting by recurrent neural networks with spiral structures of layers
- PDF / 2,701,072 Bytes
- 13 Pages / 595.276 x 790.866 pts Page_size
- 85 Downloads / 154 Views
(0123456789().,-volV)(0123456789(). ,- volV)
ORIGINAL ARTICLE
Urban traffic flows forecasting by recurrent neural networks with spiral structures of layers Vasiliy Osipov1 • Victor Nikiforov1 • Nataly Zhukova1 • Dmitriy Miloserdov1 Received: 15 May 2019 / Accepted: 6 March 2020 Ó Springer-Verlag London Ltd., part of Springer Nature 2020
Abstract The problem of neural network forecasting of processes with changing laws of their behavior and imperfection of timeseries samples is considered on the example of analysis of urban traffic flows. The goal is to improve the accuracy of such forecasting. To achieve this goal, we analyze the applicability of self-learning recurrent neural networks with controlled elements and the spiral structure of layers. Based on the development and application of these neural networks, the new methods and the system implementing them are proposed. These methods, in contrast to known solutions, allow continuous training of neural networks and forecasting of processes. There is no need to interrupt training to perform forecasting. For forecasting, it is possible to continuously take into account the properties of the observed processes. In addition, improved controlling of associative recall of information from the memory of recurrent neural networks is provided to improve the accuracy of forecasting. The results of traffic flow forecasting are presented. The results are compared with estimates obtained using other methods. It is shown that the proposed methods have advantages in accuracy compared to the known solutions. The developed methods are recommended for use in advanced robotic and other intelligent systems. Keywords Recurrent neural network Associative processing Forecasting Traffic flow
1 Introduction In recent years, artificial neural networks have been developed considerably and applied in various fields [1, 2]. Among such applications, neural network time-series forecasting has an important place [3]. The time series can carry information about physical, social, economic, financial, transport and other processes. To predict time series, various neural network models and methods are used [3–9]. Despite the achievements in development, the wellknown neural network models and methods remain far from being perfect. They are focused on solving narrow classes of problems, are not flexible to changes in external conditions, and do not provide enough possibilities for controlling the associative interaction of signals during their processing. This negatively affects the results of & Vasiliy Osipov [email protected]; [email protected] 1
Saint Petersburg Institute for Informatics and Automation of the Russian Academy of Sciences (SPIIRAS), 39, 14 Line, Saint Petersburg, Russia 199178
neural network forecasting. A search for new methods and models is needed to significantly expand the capabilities of this forecasting. To develop required models and methods, it is proposed to consider a problem of urban traffic flow neural network forecasting. Timely and accurate forecast
Data Loading...
