Research On Short-term Lease Forecasting Method Of Urban Public Bicycle Stations

As a kind of shared transportation,urban public bicycles have been adopted by major cities and bear the heavy responsibility of the “last mile” of urban public transportation.At present,the main problem of the urban public bicycle system is that it is difficult for users to rent a car during peak hours,and real-time monitoring cannot be solved well.Therefore,predicting the demand for bicycles in a certain period of time and performing redistribution in advance is of great significance for solving the lag of bicycle system scheduling.Based on the data of the City public bicycles and the Central Park Meteorological Monitoring Center in New York,this paper forecasts the short-term rental of urban public bicycle site clusters.Firstly,the public bicycle rentals of individual stations,site clusters and the whole city are analyzed separately.The results show that it is difficult to predict the rental amount of each bicycle station separately and the value performance is low,and the rental amount of the predicted site clusters is more practical.And K-means,K-medoids clustering algorithm and exponential smoothing model,autoregressive moving average model,RNN,LSTM prediction model were compared and analyzed.A city public bicycle prediction model based on long and short time memory network model(LSTM)is proposed.Firstly,the model uses the K-medoids algorithm to cluster all the sites of the City bicycle system in New York.Two representative site clusters are selected,and the bicycle data of the two clusters are input into the established LSTM model for prediction.Finally,the paired test and the real value are compared and analyzed,and the root mean square error and error rate are calculated.The average correlation coefficient,root mean square error and error rate are 0.9425,0.298 and 0.296,respectively.The prediction effect is good.Finally,based on the LSTM prediction model,a prediction model of urban public bicycle based on hybrid model is proposed.The model is divided into three parts.The first part proposes a two-factor clustering algorithm based on the geographic location and historical transition model of the bicycle station to cluster the stations.The second part is built by considering the factors of time,weather,temperature and wind speed.The LSTM model is used to predict the rental of public bicycles in the whole city.The third part proposes an inference model based on the coefficient of variation function.According to the model,the proportion of the total occupancy of each cluster is calculated,and then the bicycle rental amount of each cluster is predicted..Finally,the predicted and true values are analyzed and calculated,and the average correlation coefficient,root mean square error and error rate are 0.956,0.28 and 0.274,respectively.The prediction effect is better than the LSTM model.