A Study Of Fluctuation Analysis And Short-term Prediction Of Bus Passenger Flow Volume

Intelligent transportation is an important component of the smart city construction.With its large capacity,low cost and high environmental benefits,public transportation is becoming the focus of attention in the development of intelligent transportation.Bus passenger flow is basic data and important basis in bus scheduling and efficient allocation of public transportation resources.The scheduling of operations has a direct effect on the productivity,economic efficiency and service quality of public enterprise.However,changes in the bus daily passenger flow present strong non-linearity and non-stationary,which makes it difficult to predict daily passenger flow due to the impact of weather,holidays and other factors.Therefore,it is of great significance for bus scheduling to master the characteristics and rules of bus passenger flow and propose an efficient prediction method.For better to deal with the complexity of bus passenger flow fluctuation and difficulty in passenger flow prediction in traffic dispatch field,this paper makes the following research.Aiming at the complexity of bus passenger flow fluctuation,passenger flow is analyzed by statistical analysis methods combining with it uneven distribution in time and space.Based on the advantages of EEMD in processing the nonlinear,non-stationary signal,passenger flow sequence is decomposed by EEMD.Then the components are reconstructed to obtain a high-frequency component,a low-frequency component and a trend component.Finally,the actual meaning of the three components represented is explained and illustrated.Based on the fluctuation analysis,short-term passenger flow is predicted by BP,RBF and ELM neural networks.By case studies,this paper details the parameters setting of three neural networks,and compares their predictive performance.On the basis of the analysis by EEMD and good performance of ELM neural network in the short-term passenger flow forecast,the combination method based EEMD-ELM is proposed.First,the daily passenger flow sequence is decomposed by EEMD to obtain IMFs and residual trend components on behalf of local characteristic scale of data itself.Then extreme learning machine(ELM)neural network forecasting model is set up according to the average period for each component.Finally,all the components' forecasting result is composed to get the forecast value of the original sequence.The effectiveness of EEMD-ELM prediction model is tested based on the actual traffic data of a bus line in Xi'an and the comparison with the single prediction method.