Optimization Of Bike Sharing Scheduling Problem Based On Data Driven Methodology

In recent years,with the accelerated development of urbanization in China,problems such as traffic congestion and environmental pollution are also prominent.In order to solve the traffic congestion and achieve the goal of low-carbon travel,there is a pile-type bike sharing,that is,a public bicycle rises in China,which relieves traffic pressure to some degree.Along with the advent of the "Internet plus" and "sharing economy" eras,the pile-free bike sharing has gradually replaced the role of pile-type bike sharing with the convenience and flexibility of use,and has been favored by many users.As a low-carbon and environmentally friendly green mode of travel,bike sharing plays an important role in solving the problem of “last mile”.However,as major bike sharing manufacturers rush to the market,besides the government has not yet issued a reasonable and detailed regulatory system,many bike sharing companies have already died.There are also some problems in the development of bike sharing.For example,the fault-sharing bikes are “stacked” and haven't to be recycled in time;unbalanced supply and demand of bike sharing at different demand points;unreasonable scheduling of bike sharing at different demand points.This paper carries out the following research on the above issues:It is known that the development of bike sharing in the United States is very successful,but there are many problems in China.First,this paper established the difference between pile-type bike sharing in America and pile-free bike sharing in China.Second,the user's riding rules are analyzed and compared based on the travel data of the bike sharing.This paper aims to understand the characteristics of different types of bike sharing and the rules of the users.Third,this paper established a model of predicting the requirement of bike sharing scheduling.Then the characteristics of the demanding quantity are excavated and compared with the established SVM,RF and LR models.We get the conclusion that the demand for the weather and other dispatching demand points has a great influence on the demand of the target demand point.In addition,this paper also establishes a fault-sharing bicycle prediction model based on Cox risk regression,and proposes five covariates that affect fault of bike sharing.Finally,we established a bike sharing scheduling optimization model with time windows constraint and considering repair or recycle bikes.Furthermore,this model aims to reduce the dispatching cost and improve the user satisfaction.We also improved a hybrid genetic algorithm with simulated annealing(HGAS)to solve the model.The results show that the algorithm designed in this paper has certain validity for solving the model,and the prediction model designed in this paper and the scheduling model has certain guiding significance for solving the scheduling problem of bike sharing.In a word,this paper expands the theory and method of reverse logistics in bike sharing economy and enriches the research on the bike sharing scheduling problem.Moreover,this paper will provide theoretical reference for further research.It is also helpful for enterprises to repair the slightly faulty bikes in time,to recover the severely bikes,in order to ensure the reuse of resources and save the cost of enterprises.