Prediction Method Of Public Bicycle Travel Demand Based On Bicycle Travel Chain

The public bicycle system is the product of the city's advocacy of "green traffic" and can effectively solve the "last mile" problem of travel and meet residents' short-distance travel needs.The reasonable forecast of public bicycle travel demand is very important to improve the utilization of the system.This paper relies on smart IC card data of Nanjing's main urban public bicycle system to define a series of travel records completed by a user in a day using public bicycles as a public bicycle travel chain.and predicts the demand for different travel purpose of public bicycle stations.Firstly,a method of public bicycle trip chain extraction based on smart card data is proposed.Based on the characteristics of public bicycle trips,combined with the smart card data structure,a public bicycle trip chain extraction method is implemented based on R language.Secondly,according to the structure of Nanjing public bicycle IC card data,the gender distribution and age distribution characteristics of Nanjing are analyzed.Then the visualization method of public bicycle trip chain is proposed.The application of second-order Bezier curve and GIS technology to achieve public bicycle trip visual analysis of trip chains.Furthermore,the continuous hidden hidden Markov model is used to identify the trip purpose of public bicycle trip chain.According to the time-space distribution of different types of trip chain,combined with IC card data,personal data of travelers and land use data,apply the unsupervised learning Hidden Markov Chain Model to identify the trip purpose of public bicycle trip chain.Finally,based on the identification of trip purpose of public bicycle trip chains,the total demand of different trip purpose is obtained.And a time series forecasting model based on long short-term memory neural network is established.Using the different land use of public bicycle site as an example,the application effect of the public bike trip chain demand forecasting model is trained and tested.