Research On Short-term Demand Forecasting And Scheduling Problem Of Bike Sharing Around Urban Rail Transit Station

As a new service mode for the integration of Mobile Internet and rental bike system,bike-sharing system is a typical format for sharing economy.With no fixed lock pile,bike sharing system adopts mobile payment technology.All of these make bike sharing system become more flexible and convenient,and hence,it attracts a lot of travelers.Bike sharing system not only greatly facilitates daily short-distance travel of the citizens but also promotes the city's Green transportation development.However,lots of sharing bikes are put on the market disorderly has caused a series of problems such as numerous bikes occupying road resources illegally,unreasonable scheduling and uneven peak passenger flow,which have brought a serious negative impact on urban traffic order and city image.These problems are urgently to be resolved.Based on the data of Mobike in Beijing for one week,this thesis makes a research on the sharing bikes around urban rail transit station.After studying actual background of bike sharing system and current researches at home and abroad,this thesis analyzes the demand of bike sharing system and discusses the short-term demand forecasting methods and scheduling methods of sharing bikes.Then the new scheduling model of sharing bikes is constructed combining nighttime static scheduling and daytime dynamic scheduling of morning peaks hours.Besides,by analyzing the case,this thesis proves that the bike sharing system has higher service level and becomes more balanced using the new scheduling model than the bike sharing system before scheduling.And this can provide reference for scheduling of bike sharing system.The main research of the thesis includes:(1)According to the relevant literature at home and abroad and operation status,the thesis summarizes the current situation and existing problems of scheduling problem of bike sharing system and determines to choose to study sharing bikes around the urban rail transit station according to the characteristics of sharing bikes.(2)The concept and function of bike sharing system are studied.And the scheduling status of bike sharing system is also studied in detail.Based on the data of Mobike in Beijing for one week,the travel demand and travel rules of sharing bicycles are analyzed.Besides,the thesis focuses on analyzing the differences of travel demand and travel rules between the sharing bicycles around the urban rail transit station and the sharing bicycles not around the urban rail transit station.From the data,it is verified that choosing to study the sharing of bicycles around the urban rail transit station is scientific.(3)Discuss the characteristics of short-term demand forecasting methods of sharing bikes and study its principle and advantages and disadvantages of short-term demand forecasting methods in detail,including parameter methods(ARMA)and non-parametric methods(KNN,SVR,BP neural network,Random forest,Xgboost).The effectiveness of the method is verified by analyzing the case.And the thesis selects the Xgboost algorithm,which has the best effect,to predict the short-term demand of sharing bikes.(4)The scheduling problem of sharing bikes is summarized,and the new scheduling model of sharing bikes is constructed combining nighttime static scheduling and daytime dynamic scheduling of morning peaks hours is built.Besides,the thesis proposes a scheduling region partitioning method combining clustering analysis and association rules.Through twice clustering,the urban rail transit stations are divided into different partition.Then combining with the characteristics of the operating mode of bike sharing system,the thesis constructs the scheduling model.The nighttime scheduling aims to pay the lowest transportation cost,and the daytime scheduling aims to meet the demand as much as possible.The most important thing is that the nighttime demand allocation is connected to the daytime scheduling.Finally,the solving algorithm of model is analyzed and a heuristic algorithm is designed to solve the model.(5)Based on the case study of Mobike in Beijing,the thesis classifies the urban rail transit stations and divide the urban rail transit stations into different partition.And the short-term demand of sharing bikes is also predicted through Xgboost algorithm.Moreover,the thesis uses the Gaode map API to obtain the actual travel time between stations and solve the model by using python programming software and Matlab programming software.It is found that service level of bike sharing system has been significantly improved than the bike sharing system before scheduling,and the bike sharing system has reached equilibrium.