| With the rapid development of the economy,ride hailing has become one of the important modes of transportation,meeting people’s requirements for travel quality,especially addressing the travel needs of residents in areas not covered by public transportation.However,due to the uncertainty of residents’ travel and ride hailing,there is an imbalance between supply and demand between the two.With the prevalence of ride hailing,a large amount of ride hailing trajectory data is generated,recording the changes in the demand for ride hailing by residents with travel requirements.By using advanced data analysis methods to deeply mine these data,the trend and pattern of the demand for ride hailing can be obtained.By analyzing the traffic data of ride hailing,we conducted in-depth research on the demand characteristics of ride hailing,and screened the factors that affect the demand for ride hailing.Develop a reasonable location strategy for popular demand points for ride hailing,and use the selected popular demand points as the central node.Divide the research area based on the Tyson polygon,and achieve clustering of ride hailing area data.On this basis,a traffic data model was used for traffic feature analysis,and a demand prediction model for ride hailing areas was constructed using deep learning theory.The model was validated and analyzed based on the ride hailing order data in Chengdu.The results of this study are of great significance for improving the operation and scheduling level of ride hailing platforms and alleviating supply-demand conflicts.The project mainly completed the following aspects of work:(1)Analysis of the characteristics of ride hailing demand and determination of related influencing factorsFirstly,clarify the definition of the relevant issues being studied and preprocess the data for subsequent analysis.From a time perspective,the study analyzes the time characteristics of ride hailing demand based on different week attributes and different time periods.At the same time,from a spatial perspective,the spatial distribution characteristics of ride hailing demand were studied and analyzed,including different weekday attributes,travel demand at different time periods,and spatial distribution of passenger carrying areas.In order to determine the input characteristic variables of the online car hailing demand forecasting model,the Spearman’s rank correlation coefficient is used to select the characteristics of the factors that may affect the travel demand of online car hailing.Finally,the input characteristic variables of the ride hailing demand prediction model were determined for accurate demand prediction.(2)Division of popular areas for ride hailing demand considering ride hailing demand data and point of interest dataBy analyzing the spatiotemporal distribution characteristics of ride hailing demand,this study explores the correlation between urban related interest point data and ride hailing demand point data.Based on the K-means clustering algorithm,a reasonable strategy for selecting popular ride hailing demand points is developed.This site selection strategy will select the popular demand points for ride hailing as the central node of the Tyson polygon,divide the studied area,and further cluster the ride hailing demand data regions to obtain ride hailing area demand data.The purpose of these works is to provide a reliable data foundation for subsequent research and support the development of effective demand prediction models for ride hailing.(3)Demand prediction for ride hailing areas based on deep learningThe demand forecasting of online car hailing area is mainly aimed at the spatiotemporal demand tensor of online car hailing area,and a method of demand forecasting of online car hailing area based on the convolution short-term memory neural network model optimized by the sparrow search algorithm is proposed.The method first uses the residual network(Res-Net)to extract and identify the characteristics of the demand data of the network car hailing area,then uses the Convolutional Short Term Memory Neural Network(Conv-LSTM)to predict the traffic flow of the regional network car hailing,and finally uses the Sparrow Search Algorithm(SSA)to optimize the network hyperparameter and network structure of the model.Selecting a dataset of ride hailing trajectories for model validation,the validation results show that the model can effectively ensure prediction accuracy and can be used to effectively predict the demand for ride hailing areas. |
