Analysis Of Station Level Data And Trip Data In Bicycle Sharing System

With the emergence of low carbon society,bicycle sharing systems(BSS)have been es-tablished in many cities worldwide.An efficient system should not only minimize operating costs,but also ensure high-quality services.So we need to acquire a good knowledge of the system data.The system data can be classified as station level data and trip origin-destination data.1.Using station level data to estimate Poisson arrival rate.First,CU-KS and Lewis-KS tests are conducted to test whether it can be considered as a Poisson process in each time period.Then we use simple mean model,piecewise-linear model,HoltWinters model in time series,FE and ME models in panel data models and LSTM model in deep learning to estimate the Poisson arrival rate of each station in each time period.Applied to Boston hubway station level data,the best-fit model is LSTM model.When the arrival rate is large,the best model for prediction is HoltWinters model.Otherwise,the best model is simple mean model.2.Using trip origin-destination data to estimate trip distributions.A comprehensive form of double constrained gravity model is established to calculate trip distribution between each station.A linear combination of users' subscription type,gender and age is also taken into account.By calculating undetermined parameters in the model,we can get results.Applied to Boston hubway trip origin-destination data,gravity model fits better than simple mean model.Through modeling and analysis of the system data,we can know interstation and intrasta-tion changes in each time period,and predict the bicycle distribution in the next period.Thus the service and management efficiency of the system can be improved.