| The emergence of bike sharing has met people’s short-distance travel needs and effectively filled the lack of travel range of public transportation.But it has also brought about imbalances in supply and demand,especially in the morning and evening peak subway station areas.The thesis mainly studies the demand forecast of bike sharing in subway station areas.It is the phenomenon for residents to use bike sharing by investigation and analyze.Combining the peculiarity of the site area,determine the main factors affecting the use of bike sharing and predict the demand for bike sharing based on the ridge regression.The main work of the thesis is as follows:(1)Research scope and data processing of bike sharing.The thesis takes Ji Luo Road Station as the site case,selects the research scope and divides the traffic districts.Using Python and ArcGIS to obtain and processe separately rail transit Site data,road network data,and POI data.(2)Investigation and characteristic analysis of the use of bike sharing.The investigation of residents’ intentions to use bike sharing for travel and the analysis of the questionnaire survey results by the distribution of questionnaires.It includes three parts: personal basic information,personal travel information and bicycle usage information.The thesis also investigates the period change of bike sharing in the site area,which analyzes the distribution of bike sharing on weekdays and holidays from three perspectives: site overview,traffic area,and vehicle category with the amount of change.(3)Established an ensemble learning model of factors influencing the use of bike sharing.First,using the bagging algorithm and boosting algorithm in ensemble learning to establish the model of influencing factors of bike sharing based on random forest and the model of influencing factors of bike sharing based on AdaBoost.Then,the parameters of the models were tuned to make them reach the optimal function Model,in the case of ensuring that the model error is smaller.Using the model to screen the influencing factors.Finally,selected 14 factors that affect the use of bike sharing.(4)Established a ridge regression model for bike sharing demand forecasting.Ridge regression model has better advantages in dealing with the collinearity problem among variables.Therefore,the demand prediction model of bike sharing weekdays based on ridge regression and the demand prediction model of bike sharing holidays based on ridge regression are respectively established.Then,the four-traffic area separate forecasts to find out the delivery volume required for each district on weekdays and holidays.The bike sharing demand forecast studied and applied to Ji Luo Road as a case study in this thesis,which can be provided a new forecasting method of bike sharing demand forecast based on subway station area and provide reference for reasonably determining the number of bike sharing in the station area. |
