Classification Of Urban Rail Transit Stations And Demand Forecast Of Shared Bicycles Around Them

Urban rail transit has been developing rapidly in recent years,the problem of “The last mile”between stations and destinations has been widely concerned.As an effective means to solve this problem,shared bicycles are not only environmental friendly,but also convenient and fast.However,due to the unplanned and unlimited delivery of shared bikes in the early stage of development,,it has also brought about problems such as vehicle accumulation and imbalance between supply and demand.Therefore,this article explores the short-term prediction methods of shared bicycles in different types of rail transit stations,and also lays the foundation for subsequent vehicle deployment.Firstly,Take the rail transit development situation and the characteristic as the breakthrough point,this article compares the influencing factors of multiple rail transit connection methods by searching related literature,web pages and other materials,analyzes the development process and advantages of shared bicycle connection rail transit,and defines The acceptable distance for the transition between rail transit stations and shared bicycles is walking as the transition,and the reasonable range of walking transition distance for rail transit interchanges with shared bicycles is proposed to be 200 meters.Secondly,this article takes the Mobike bicycle order data in Beijing in May 2017 as the research object.By using the python language and Geo Hash algorithm to select and preprocess the data,the existing data is filtered and cleaned,and an effective data set is selected.Extract and analyze spatial data,and display the spatial and temporal distribution characteristics and distribution influencing factors of shared bicycles through heat maps,time distribution diagrams and other visual charts.Combined with the weather,temperature,wind speed and other factors at the time,the characteristics of the use of shared bicycles under different temporal and spatial distributions are obtained.Then,a variety of clustering methods and principles are compared,and the K-means clustering algorithm is finally selected to cluster the track stations.Taking the hourly borrowing and repayment of shared bicycles within 200 meters of the site as the clustering variable of the site,the sites are clustered into 5 categories.Detailed analysis,description and explanation of the average hourly rental volume of 5 types of sites.A set of clustering methods that can be analyzed by quantitative indicators are proposed,which lays a research foundation for the subsequent demand forecasting of various rail stations.Finally,the random forest and lasso regression algorithms are used to predict the demand of shared bicycles around the rail stations in a single and combined model.The demand forecasts for the five types of stations are carried out respectively,and the comparison and analysis with the actual data are carried out.Whether the final prediction result of this model is good or not,it is found that the combined model has the highest accuracy,and a stable shared bicycle demand prediction model is obtained.