Bus Passenger Flow Prediction And The Optimization Of Bus Operation Period Division Based On Multi-source Data

Bus transport system is a relatively open system,numerous and complicated factors influence the operation of public transportation system,including the internal factors such as bus passenger flow fluctuation,fleet vehicle structure,scheduling plan implementation,and the external influence factors such as working days holiday properties,temperature,weather,road traffic and other uncontrollable factors.All these factors affect the operation process of public transport in different ways.Public transport managers need to constantly optimize their operation strategies according to the changes of these factors.The previous public transport management was not very delicacy.The proposal and change process of public transport management schemes were generally driven by experts' experience rather than real time data.In recent years,with the improvement of infrastructure,tens of millions of passenger travel records and bus driving records can be collected in large cities every day.More and more bus data can serve as the key reference for bus management department.For the urban bus routes with huge passenger flow,complex fluctuation of passenger flow and complex road conditions,it is difficult to accurately implement the operating plan.The influence of the factors that impacts the bus system should be minimized to provide a stable environmental variables for the bus system robustness optimization and real-time control process.To achieve this,there are two main strategy for this purpose,one is to predict these variables more accurate and reasonable,providing a set of optimal parameters settings for operation planning,the second is by adopting the idea of divide and rule,on the basis of the optimal parameters,divide the operation hours of the bus route into several reasonable time intervals,the operation parameters of each interval is accurate and stable,with similar supply and demand of transportation capacity,so that managers could choose the optimal operating strategy.In this paper,four topic is proposed and are studied in depth: The passenger flow prediction of a bus route,the peak-load passenger flow prediction,the time of day interval partition problem,the the time of day interval partition and fleet configuration.In order to predict the passenger flow accurately,this paper proposes a new method to predict the passenger flow which takes into account the passenger behavior pattern.The traditional prediction method takes the total volume of the passenger as the input of their prediction model,ignoring the different types of passengers‘ unique respond mode to the change of the external influencing factors,ignoring the time-changing law differences of different types of passenger volume fluctuations.To overcome these defect,this paper proposes a set of bus passenger classification method and demonstrates the necessity of prediction,with the purpose of improving the prediction precision.Based on the previous analysis,a hybrid machine learning model of passenger forecast is proposed,the response mode of different types of passengers to different external factors and time-varying law of different types of passenger are all fully considered.The forecast precision of passenger flow forecast model is improved greatly.Based on the analysis of the complex influence of the peak-load passenger flow prediction error on the operational risk cost,this paper proposes a model of the peak-load passenger flow prediction based on the newsboy model.Traditional peak-load passenger flow prediction are all abstracted to a simple numerical prediction problem,ignoring the actual impact of the peakload passenger flow prediction error on the cost of operational risk.According to the actual bus operating parameters,this paper calculates the operational risk brought by peak-load passenger flow prediction error,and replacing the loss function of traditional prediction model with operational risk cost.In addition,the Shepard interpolation prediction model has high prediction stability and accuracy,so this model is selected and improved in this paper to make it applicable to the prediction problem based on the newsboy model proposed in this paper,and the passenger flow prediction model with the best prediction performance was obtained.As for the problem of time interval partition problem,based on the analysis of the purpose of time interval partition,this paper proposes a time interval partition method that takes the similarity of capacity demand in consideration.The current research on the time interval partition problem is preliminary,and the existing method of time interval partition is generally aimed to minimize the difference of single operation parameters(such as passenger demand or bus travel time)within the period.Such studies have ignored the cooperation of these operating parameters on the bus operation optimization work,this paper deeply analyses the deficiency of these studies,and proposed an internal capacity similarity based time interval partition method.Based on the previous analysis,this paper proposed a time interval partition method based on the similarity of capacity demand.By using the multi-source bus data,a frame work of operating cost is established.By using this frame work,an optimization model with purpose of minimizing the fleet-time cost is established.The result shows that the time interval partition scheme given by this model is efficient.Through the in-depth study of the time interval partition problem,the author finds that the time interval partition problem is closely related to the fleet configuration problem,and on this basis,this paper proposes a collaborative optimization model for time interval partition and fleet configuration scheme.Through the research on time interval partition problems,the author found that the capacity demand is a reasonable index for time interval partition,but fleet capacity is also a complex concept,it contains two important sub-problems: the vehicle choice problem,fleet size optimization problem.When these two subproblems are considered into bus time interval partition problem,the problem is complicated.Based on previous analysis,this paper innovatively proposes a collaborative optimization on time interval partition and fleet configuration.In the process of cost calculation,a large number of multi-source bus data is still used to calculate the all-day environmental variables.In addition,the newsboy model is also conducted to calculation of the operational risks of different fleet configurations.On this basis,with the purpose of minimizing the all-day operation cost,this paper establishes an optimization model for the collaborative optimization of time interval partition and fleet configuration,and obtains a good time interval partition scheme and fleet configuration optimization scheme.