Analysis Of Cycling Characteristics And Prediction Of Rental And Return Quantity Of Dockless Shared Bicycles In Nanjing

Shared bicycles,as a new urban transportation method developed rapidly in recent years,has been welcomed by people because of its can be retrieved or returned at any time or anywhere?flexible and convenient,which has changed the way of residents travel.However,during its development,problems such as imbalanced supply and demand,random stop and other problems have also occurred,which seriously affected the appearance of the city and even the unblocked of the urban transportation system.In order to better play the complementary role of shared bicycles in urban transportation and improve the residents travel experience,it is necessary to standardize the management of shared bicycles based on a deep understanding of the characteristics of shared bicycles.Therefore,this paper takes shared bicycle data as the research data source,and combines the statistical analysis method,GIS spatial analysis method,swarm intelligence optimization algorithm,and deep learning algorithm to do the following research.First,analyze the basic characteristics of shared bicycles.Explored the hotspot periods of bicycle use,and identified the morning and evening peaks of bicycle use.Shown that the use of shared bicycles is mostly less than 1,000 meters,and the duration is less than 30 minutes,mainly serving short-distance travel of residents.Similar to the urban transportation system,the morning rush hour and evening rush hour will be respectively generated at 7: 00-9: 00 and 17: 00-19: 00.Use the kernel density analysis and quad-segmentation to identify hotspots for cycling and understand the spatial characteristics of cycling.Bicycle use forms different hotspot areas in the morning and evening peaks,among which the hotspot areas for bicycle use and aggregation are most likely to be formed near the subway station and Business center.Second,combined with the urban function type region,analyze the periodicity of cycling.Use the time series analysis method to analyze the cycle characteristics of shared bicycles.By analyzing the rental and return of bicycles in different types of urban interest surfaces,the characteristics of different periodic changes corresponding to different types of city area of interest are obtained,and their corresponding travel behaviors are analyzed.Studies have shown that the morning rush hour in commercial building areas is more used as the destination point of travel,and the evening rush hour is used as the origin point for travel;correspondingly,the residential area is used as the origin point for travel in the morning rush hour and the evening peak is used as the destination point of travel,verifies the commuting needs of residents.There is not much difference in bicycle usage during the daytime in scenic spots,indicating that residents arrive or leave scenic spots in more scattered time.The use number of bicycles in science and education cultural areas in the morning peak period is significantly greater than that in the evening peak period.Change law of bicycle cycling number in other service areas is similar to that in commercial buildings.Comparing workdays with rest days,the use of all types of urban interest surfaces on workdays is more reflective of morning and evening peaks than on rest days,that is,there is a large difference in use number of bicycles at different times during the workday.Third,a deep learning algorithm and a swarm intelligence optimization algorithm are used to build a model for prediction of rental and return quantity of dockless shared bicycles(QPSO-LSTM).The Long-Short Term Memory(LSTM)which suitable for the time series prediction is selected as the core algorithm for prediction.In order to solve the problem that the number of hyperparameters in the LSTM is large,the group intelligent optimization algorithm is used to optimize the hyperparameters,which can quickly determine the appropriate hyperparameter combination.The application and accuracy verification of the model for predicting the number of shared bicycles rented back are applied.The results show that QPSO-LSTM model can effectively learn the cycle law of the change of bicycle rental and return quantity,and the prediction result of bicycle rental and return quantity is more accurate.Finally,the QPSO-LSTM model constructed in this paper is compared with artificial neural networks,deep learning algorithms,and time series prediction models.The results show that QPSO-LSTM model has improved accuracy compared with previous models.