Research On Short-term Traffic Flow Forecasting Based On Bagging Integrated BP Neural Network

At present,intelligent transportation has become a new idea and technology to effectively alleviate urban traffic congestion.Real-time and accurate traffic flow prediction plays an important role in intelligent transpo rtation.Effective and reasonable short-term traffic flow prediction provides accurate and real-time traffic change information for traffic managers and travelers,and provides accurate basis for travel choice.However,short-time traffic flow is characterized by random volatility and complex nonlinearity,which makes the prediction accuracy and time of existing methods difficult to meet the requirements of actual intelligent transportation.Therefore,it is necessary to carry out further research work.Thi s paper studies the short-time prediction method of traffic flow from three aspects: preprocessing of traffic flow data,(Back-Propagation,BP),improvement and integration of BP neural network.The main work is as follows:1.Traffic flow data preprocessing is studied.In view of the strong correlation between the traffic flow data acquisition process and the external environmental noise and the time series,a BP neural network short-time traffic flow prediction method based on wavelet and multidimensional reconstruction is proposed.Firstly,the heuristic wavelet noise reduction method is used to deal with the original traffic flow data to eliminate the noise in the data.Secondly,the multi-dimensional phase space reconstruction of the traffic flow data is carried out by C-C method,and the multi-dimensional variation characteristics of the traffic flow are fully explored.Moreover,a multi-dimensional BP neural network is constructed for short-term traffic flow prediction,so as to further accurately refle ct the fluctuation of traffic flow.2.Establishment of prediction model based on IBA-BP neural network.In view of the short comings of BP neural network in traffic flow prediction,such as poor accuracy and sensitive weight setting,an improved(Bat-Algorithm,BA)was proposed to optimize the short-time traffic flow prediction method of BP neural network.Firstly,adaptive inertia weight and acceleration factor are introduced to optimize the original bat algorithm to improve its convergence speed and prec ision.Secondly,the(Improved Bat Algorithm,IBA)was used to optimize the weight and threshold parameters of BP neural network and to construct the IBA-BP model for short-term traffic flow prediction,so as to improve the accuracy of the prediction model.3.Integrated IBA-BP traffic flow prediction model based on Bagging method.In view of the poor classification effect,low generalization ability and low prediction accuracy of a single BP neural network,a short-time traffic flow prediction method integrating Bagging with BP neural network is proposed.firstly,collected data will be preprocessing,and divided into training set and test set,and then through the Bootstrap method to training set were randomly divided into sereval training data subsets,some training subsets are used to model training by using BP neural network,finally to test the model with test data sets,and the output results are weighted average to get the final result.Multiple prediction results determine the prediction results,greatly improving the stability of the model and prediction accuracy.