Research On Demand Forecasting And Scheduling Method Of Public Bicycle Sites

The emergence of urban public bicycles has effectively solved the problem of short-distance travel “last mile” and has become an indispensable link in the public transportation system.However,at the same time of its rapid development,due to the tidal traffic and the untimely dispatching of vehicles,the imbalance between supply and demand of public bicycle system sites has occurred,which greatly affects the development of the system and the experience of travel users.How to accurately predict the travel demand of each station in the system and carry out reasonable vehicle scheduling on this basis is of great significance to solve the lag of scheduling.Therefore,it is essential to improve the scheduling mechanism and realize the “balance of the bike with rent and availability”.This paper analyzes the influencing factors of urban public bicycle travel,predicts the travel demand of the station and and proposes the system’s vehicle dispatching mechanism,which provides an important theoretical basis for the sustainable development of the urban public bicycle system.Firstly,we introduce the research status of urban public bicycle station planning,demand forecasting and vehicle scheduling problems at home and abroad,and analyzes the current imbalance of demand and unscheduled scheduling of public bicycle system development.Data mining is based on public bicycle travel data and weather data in the Bay Area of the United States.The influence of meteorological conditions(temperature,humidity,wind speed,cloud cover,weather)on the travel demand of the site was analyzed and the correlation analysis was carried out.The travel demand of the site was measured from the time scale(month,week,day,hour).In addition,the correlation between the predicted site and the related site is analyzed to grasp the travel characteristics of the system site.Secondly,based on the analysis of the factors affecting the travel demand of the system,clustering algorithm is used to cluster the sites in time and space to obtain the relevant site clusters of the site.Next,the logarithmic optimization method is used to transform the anomaly data,reduce the influence of the anomaly data on the prediction model,and construct a demand forecasting model by using the random forest with better generalization performance.Different prediction models are established for different site clusters,so that the model is more targeted to the data in the same site cluster,thereby improving the prediction accuracy of the prediction model.Finally,we introduce the definition and classification of vehicle scheduling problems,and analysis the problems existing in the current public bicycle system vehicle scheduling.Considering the randomness of user travel demand,combined with the above prediction analysis results,a vehicle scheduling method based on price incentive mechanism is proposed to realize the supply and demand balance of public bicycle system sites.While improving service levels,the need for system dispatchers to relocate vehicles is reduced,thereby saving system operating costs.