Prediction Method Of The Demand Of Public Bicycle Based On The Associated Rental Stations

As a vehicle, public bicycle system can solve problem of the "last mile" and effectively promote the overall service level of urban public transport. Nevertheless, in the rapid development of the public bicycle system, it often appears that there are no bicycles or vacancies in the rental stations. As the demand of bicycles at different stations in different periods are imbalanced, the bicycles in the stations need to be re-balanced manually to keep its balance (that is, to meet the needs of users). However, only forecasting accurately the future demand of the rental points, can it ensure that the amount of scheduling bicycles is exact. Therefore, this thesis presents a prediction model based on the associated rental stations, which could predict the number of bicycles rent/retumed from each station in the future period. It will lay the foundation for the calculation of the bicycle dispatching model.Firstly, taking Hangzhou as an example, it needs to deal with the large-scale data of public bicycle system in Hangzhou. At the same time,it is necessary to analyze the characteristics of the public bicycle system, including travel characteristics and demand characteristics at leasing points. According to area of the rental station (such as residential area), leasing point should be classified. Direct at the same type of leasing points and the different types of leasing points, influence factors and demanding number of bicycle are discussed.Secondly, in the light of the dynamic nature of process of rented and returned bicycles,as well as the mutual influence between lease points, a hierarchical clustering algorithm is proposed. By the location of leasing points and traffic between leasing points, we can get relevant rental station clusters. Combined with the characteristics of the VAR model,finally it will establish prediction model of demand of public bicycle based on the associated rental stations.Finally, the data of four public bicycle systems(Hangzhou, Wenzhou, New York and Washington) will be used to conduct case studies, as well as comparing with baseline method, historical average method and ARIMA model. The results show that prediction accuracy of the model, which is based on the associated rental stations, not only higher than other methods, but also can be applied to public bicycle systems in different cities.