| With the rapid development of domestic tourism,the problems such as unbalanced distribution of tourist flow and the mismatch between tourism traffic demand and tourism traffic infrastructure gradually become prominent.In order to ensure the steady development of tourism,it’s vital to grasp the demand of tourism transportation and make overall planning of tourism transportation.With the popularity of ride-sharing,car-hailing data with wide coverage and rich information can break the limitations of survey data and more comprehensively reflect the characteristics of tourism traffic.In this context,this paper explores tourist identification method in scenic spot based on car-hailing data,and uses the identified tourist flow to carry out the research on prediction of urban tourist demand and distribution.The main contents and conclusions include the following three aspects:(1)Based on the OD trip data of car-hailing,identify tourists of scenic spot.Firstly,by referring to the definition of tourists,this paper expounds the connotation of car-hailing tourists,and divides the scenic spot into three types:H,M and L according to the POI density of the environment where the scenic spot is located,and analyzes the distribution characteristics of tourists around the scenic spot.And on this basis the scenic spot geometry is constructed through the scenic spot registration point and entrance and exit,each trip is matched with the scenic spot,and tourist trips are extracted from the car-hailing database.Finally,it verifies and analyzes the distribution characteristic of tourists,and finds that Numbers of tourists in urban scenic spot is higher than that in suburban scenic spot.Within a week,the average daily tourists on weekends is higher than that on weekdays.In one day,the time-varying graph of the average tourists shows double peaks.(2)Based on the principle of LSTM model,the prediction model of hourly granularity of car-hailing tourists in scenic spot is established.According to the time characteristics of periodicity and dynamic continuity,the feature set was constructed and the importance of feature variables was sorted by the random forest algorithm.Then the variation trend of variable significance is combined with the variation trend of model performance to determine the optimal feature subset as the input feature set of the model.In the application of the Summer Palace as an example,the model R~2 reaches92.8%,and the prediction can reflect the trend of tourists,and also reflect the growth of tourists during holidays.In the peak period of tourist flow,there are many tourists and the fluctuation is high,which needs to focus on and strengthen management.(3)Based on the spatial distribution characteristics of tourist flow sources,Kmeans++clustering algorithm is applied to divide tourist traffic zone.After that,the influencing factors of tourism distribution were analyzed from the aspects of attraction,zone travel power and traffic impedance,and a tourism distribution model based on the improved gravity model was established.Stepwise regression and ridge regression were used to calibrate the parameters.The stepwise regression could guarantee the estimation ability of the model.Ridge regression was used to interpret the variables and analyze the influence of various factors on tourism distribution.In the application of 4A and 5A scenic spots in Beijing,it is found that the tourists are mainly distributed between the scenic spots with high tourist flow and the traffic zones close to the scenic spots.The attraction of scenic spots is the main driving factor of tourist flow,and the distance is the main obstacle factor of travel.The results can be used as reference for the operation of urban scenic spots and traffic management. |
