Research On Short Term Traffic Flow Prediction Based On Multi-Scale Wavelet Decomposition And Deep Learning

Under the background of the continuous improvement of urbanization development level,the increase of people's travel demand has led to an increase in the number of urban motor vehicles year by year,which brings a series of problems such as traffic congestion to the urban traffic system.Intelligent transportation system has been widely used as the key to alleviate these problems.In recent years,with the development of traffic information technology,intelligent transportation also needs to improve its related technology to promote traffic intelligence urgently.Traffic flow prediction,as a basic technology in intelligent transportation system,can provide real-time and dynamic guidance information for formulating traffic management and guidance measures.Therefore,short-term traffic flow prediction of urban roads is of great significance in easing traffic congestion and improving the capacity of urban road networks.Based on the actual data and the characteristics of traffic flow,this paper proposes a multi-scale particle swarm optimizing deep belief network model for short-term traffic flow prediction,which improves the accuracy of short-term traffic flow prediction.Firstly,the background and significance of the topic is determined.It is briefly summarize that the basic knowledge of short-term traffic flow and the research status of deep learning.And then,the main research contents and technical routes of this paper is given.The basic knowledge and theoretical basis including wavelet analysis,deep belief network and particle swarm optimization are introduced.The model principles of the above models and algorithms are described in detail,and the related parameters are analyzed.Secondly,based on the theoretical knowledge,a short-term traffic flow prediction model based on deep belief network is established.The framework and prediction process are given and the relevant parameters of the model are analyzed.Under the condition of analyzing the parameters of deep belief network,a short-term traffic flow prediction model with particle swarm optimization algorithm is proposed.Based on the plasticity of deep belief network,a model based on multi-scale wavelet decomposition fusing deep learning is established.Finally,the real historical data are used to verify the model.On the premise of analyzing the basic data and determining the evaluation performance index,by analyzing the influence of the number of hidden layer neurons and learning rate on the model,the prediction accuracy of different models is compared and analyzed.Finally,the short-term traffic flow prediction model based on multi-scale wavelet decomposition fusing deep learning is selected.On this basis,the accuracy of the model in prediction workday,non-workday and peak hour is also analyzed.The simulation results show that the short-term traffic flow prediction model based on multi-scale wavelet decomposition fusing deep learning has good performance.The mean absolute percentage error of the model for one week traffic flow prediction is 7.21%;the mean absolute percentage error of the working day is about 7.97%;the mean absolute percentage error of non-working day is about 10.39%;the peak hour prediction accuracy is between 95%and 96%.