Short-term Forecast Of Urban Rail Transit Passenger Flow Based On Deep Learning

Urban rail transit is an important part of urban public transport system.From the point of view of the urbanization process,urban rail transit has gradually become the preferred means of transport for travel due to its fast air speed,large throughput and safety and stability.At present,China has entered a stage of rapid economic development.Rail transit has become an important means to develop urban economy and improve residents ’ living standards.With the development of economy and the increasing demand for travel,the role of rail transit in stimulating domestic demand and promoting economic growth is more obvious.Urban rail transit,as a transportation tool vigorously developed by the first-tier and second-tier cities in China,has irreplaceable social and economic attributes.The establishment of urban rail transit system,on the one hand,can optimize and adjust the travel routes of urban residents in a timely manner,especially when there are a large number of large,such as holidays,it plays a guiding role in residents ’ travel.On the other hand,when an emergency occurs,the existence of urban rail transit systems also facilitates operators to adjust their operating lines and rescue strategies in a timely manner to ensure safe travel of urban residents and complete normal service tasks.As a public service infrastructure,urban rail transit requires substantial economic investment and advanced monitoring technology.The design of urban rail transit system can make the public enjoy convenient travel conditions while reducing travel costs.The key is to effectively improve the accuracy of short-term passenger flow prediction,and real-time monitoring and accurate prediction of passenger flow in each period of each station.This paper studies the prediction methods of urban rail transit passenger flow in different access stations.The main contents are as follows :(1)The similarity and periodicity of the passenger flow data of each station in the subway operation route are analyzed.The change of passenger flow in a week is selected for comparative feature analysis.By analyzing the change of passenger flow in a fixed period of time every day,it lays the foundation for the wavelet analysis and the establishment of the prediction model.(2)Traffic passenger flow data wavelet analysis processing.The main way to preprocess the passenger flow data is to train the original traffic passenger flow data and the traffic passenger flow data after wavelet processing,and then establish the standard LSTM-1 and LSTM-2 model structures and carry out comparative experiments to verify the effectiveness of wavelet threshold denoising..(3)Construction of LSTM model.The characteristics of various LSTM models are analyzed to establish the LSTM model in this paper.By optimizing the experimental parameters,the internal structure of LSTM hidden layer is compared and analyzed.The LSTM-1 and LSTM-2 models are trained by using data sets with different data characteristics,and then the prediction accuracy and stability of the model are verified by dividing different prediction time steps and comparing with several other comparative models.(4)Improving the construction of spatio-temporal LSTM model.The hidden layer and neuron structure of the general LSTM network are improved to fully obtain the characteristics in the time series passenger flow data.Then the LSTM-3 model is constructed and trained by inputting the traffic flow data after wavelet denoising.(5)Attention mechanism is introduced into LSTM-3 model to construct LSTM-4 model.The AI subway passenger flow data of Ali Tianchi Dasai Tianchi City are selected for case analysis,and the performance of LSTM-4 model is evaluated,and compared with ARIMA model and LSTM-3 model.The experimental results show that the proposed prediction model has better prediction effect.