Open AccessAnalysis and prediction of traffic flow based on Wavelet-BP neural network
Author(s) -
Zhuo Wang,
Lei Zhao
Publication year - 2019
Publication title -
journal of physics. conference series
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.21
H-Index - 85
eISSN - 1742-6596
pISSN - 1742-6588
DOI - 10.1088/1742-6596/1325/1/012067
Subject(s) - traffic flow (computer networking) , traffic generation model , wavelet , computer science , artificial neural network , matlab , data mining , traffic congestion reconstruction with kerner's three phase theory , simulation , real time computing , algorithm , artificial intelligence , engineering , traffic congestion , computer network , transport engineering , operating system
Through on-the-spot investigation, collected data of traffic flow from a certain section of 3rd Ring Road in Beijing, the investigation results were analyzed from the perspective of time and space, macro rules of traffic flow and the influence of incidental traffic events on roads was summarized. The MATLAB software was used to decompose traffic flow data by three-layer wavelet, and the variation characteristics of the detail coefficients were obtained by decomposing the parameters of traffic flow by wavelet. Analyzed the abrupt change phenomenon of the traffic flow parameters, so as to judge whether some traffic incident occurred and which kind of it. With the level of road traffic service, the same moment which anomalous traffic flow data produced theoretically and traffic service levels were greater than or equal to 4 levels was chosen, so as to verify whether the anomalous time point analyzed by wavelet-analysis were precise. Used MATLAB software to establish the prediction model by a neural network, and determined the various parameters, functions, training and testing samples. Coefficients of different parameters in the BP neural network were used to predict the changes of traffic flow in the future. Compared the predicted data with the actually observed traffic flow, the starting time of traffic incident and the anomalous mutation starting point of three traffic flow parameters was consistent. This research had great significance to the future intelligent traffic decision-making.