| Public bike system(PBS),the product of green development of urban transport,undertakes the service function of the "last mile" of urban public transportation.At present,the problem of public bicycles redistribution during peak hours is still the main bottleneck restricting the smart and efficient development of PBS.With the advent of the era of big data,increasingly mature data mining methods have provided new ideas for solving the problems of PBS.Based on the PBS’s historical operation data,POI data,weather and air quality data,this thesis deeply explores the spatiotemporal characteristics and influencing factors by the usage of PBS from a data-driven perspective,and then studies the methods of demand forecast and balanced scheduling of public bike station.Firstly,on the basis of preprocessing various types of data,this thesis explores the use characteristics of public bicycles from temporal and spatial perspectives.In temporal,in order to improve the efficient of analysis,the statistical clustering method is adopted to classify the bike stations into 5 categories,and then time-varying demand patterns of each category are analyzed.Spatially,the spatial data analysis methods such as Kernel density estimation,spatial autocorrelation,cold-hot spot analysis are used to explore the spatial distribution characteristics of station usage,defining OD correlation coefficient to analyze the connection strength of stations.Furthermore,a method for identifying the land use type of bike stations based on POI data is proposed,and the stations are divided into residential,traffic infrastructure,office,and commercial leisure categories.Secondly,according to the demand characteristics of stations and the applicability of forecasting methods,A fine-grained random forest prediction model with three types of influencing factors as meteorological factors,time characteristics,and location of stations as characteristic variables,and the station demand in 15 minutes、30minutes and 1 hour as target variable is constructed.The verification results of historical data show that the model proposed in this paper has high prediction accuracy.Furthermore,for the balanced scheduling of public bicycles during peak periods,a research idea that balanced scheduling is carried out according to divided regions is proposed.Through problem analysis,a model for dividing scheduling regions based on the mixed attribute of the actual route distance between stations and the OD correlation coefficient is constructed,and the AP algorithm is used to solve the model.Under the premise of dividing the scheduling area,with the minimum scheduling cost and time penalty cost as the objective,a scheduling model of scheduling vehicle path with time window constraints is constructed,and the genetic algorithm is designed to solve the model.Finally,taking the PBS of Ningbo as an example,the balanced scheduling optimization model proposed in this paper is verified.Based on the scheduling area division method,stations are divided into 28 sub-scheduling areas,and the genetic algorithm is developed in Python to design the scheduling optimization plan of PBS during the peak period.The results show that the proposed regional balanced scheduling method in this paper has certain practical value,and can achieve the balanced scheduling optimization of public bicycles during peak periods. |
