Prediction Of Traffic Flow In Large Scale Activities Based On The Combination Of ARIMA And ANN

With the development of economy,the demand for traffic travel is increasing,and the number of motor vehicles is increasing.The traditional traffic technology has been difficult to solve the serious traffic jam.Intelligent traffic has been put forward as an effective way to improve urban traffic conditions.The traffic flow prediction is the premise and key of intelligent traffic.On the other hand,more and more large scale activities have been held.During the large-scale activities,the traffic flow would be gathered rapidly in the limited space,resulting in traffic jam in the surrounding area,which brings great challenges to urban management.In this paper,the regional traffic flow prediction in large-scale activities is taken as the research object.The characteristics of regional traffic flow and its influencing factors under the were analyzed.The effective methods of regional traffic flow prediction under large-scale activities were discussed.The purpose of this study is to provide guidance for pedestrians’ travel and traffic management.The main contents of this paper are as follows:1.Taking the musical fountain in Dalian as an example,the characteristics of regional traffic flow under the large-scale activities are analyzed.After data collection,data preprocessing and traffic flow characteristics analysis,it is found that the regional traffic flow under the large activities has some typical characteristics,such as sudden increase,periodicity,linear correlation,randomness and strong nonlinearity.The sudden increase reflects a surge of regional traffic flow before and after the activities.The periodicities reflect a typical diurnal variation and seasonal variation.The linear correlation is a strong linear correlation between the time series of traffic flow.Besides,and the traffic flow at the specific time shows strong randomness and nonlinearity.2.This paper establishes the prediction model of regional traffic flow based on ARIMA model and BP model respectively.ARIMA model has good effect on short-term prediction of regional traffic flow.But it is weak for strong nonlinear processing,resulting in a large predictive errors at peak value.In theory,the BP model has the ability to approximate arbitrary nonlinear.However,the short-term prediction effect on the regional traffic flow under large activity is not good,because it is easy to fall into the local optimal solution under multi-influential factors.It indicates that a single prediction method is difficult to achieve good prediction results,due to the complexity of the actual traffic flow.3.A regional traffic flow prediction method based on the combination model ARIMA and ANN is proposed.The time-distribution statistical model of traffic flow is utilized to consider the influence of the daily-periodicity and the known factors.Then,the ARIMA model was used for predicting the linear components in the traffic flow and the BP neural network model was used to predicting the nonlinear residual components in the traffic flow.The final prediction results were obtained by combining the linear and nonlinear components.The results shows that the prediction accuracy of the combined model is is higher than that of separate ones.