Individual Travel Behavior Prediction Based On Trajectory Data

The rapid development of cities has made the phenomenon of traffic congestion and irrational resource allocation more and more prominent.How to effectively solve the traffic problems faced in urban development is a key issue in the process of urban planning.The study of human mobility can reveal the inherent laws of human mobility,and it can also help traffic managers better promote the construction of smart cities.There are also many key applications in the prediction of individual mobile behavior,such as traffic flow prediction,station passenger flow prediction,and urban development planning.In this paper,an individual Markov travel prediction model is established to predict the travel status of users,and a Long Short-term Memory Network and Origin-Destination Spatio-Temporal Graph Convolutional Network are used to predict passenger traffic at subway stations.The research contents of this paper mainly include the following aspects:1.In order to better understand individual travel behavior,this article mines user attributes and travel characteristics from raw data.According to the statistics of the travel distribution of subway users,it is found that most users prefer a certain travel mode.According to the different travel habits of users,all users are reasonably divided into three categories: commuters,non-commuters and non-home station users,and define their home,work and other types of stations.After labeling user travel records,the differences in travel distribution between users with different attributes are further analyzed.2.After understanding the travel habits of users from the group level,establishing individual travel prediction models for users can study the travel differences between different individuals.According to the user's historical travel status,first-and second-order Markov travel prediction models are established for each user.The Markov model was used to make single-step and multi-step predictions of the user's future travel status,and the prediction accuracy reached 85.82%.After predicting the travel status of each user,the future traffic of each subway station is also predicted,and the prediction result accurately reflects the real traffic.3.Precise station passenger flow prediction plays a key role in alleviating traffic congestion.Metro traffic data is a kind of time series data.Recurrent neural network can effectively capture the time correlation of the data,and overcome the problems of high calculation cost and low time resolution of individual Markov models.In order to further improve the accuracy of station traffic prediction,general and independent long-term and short-term memory network traffic prediction models for subway stations are established.The model cascades data into short-,medium-,and long-term historical passenger flows and enters the model,which can reasonably handle the temporal correlation of subway station traffic.By predicting the passenger flow of each site in the next month,the accuracy rate of the long-term memory network traffic prediction model reached 91.2%.4.Although long-term and short-term memory networks can capture the time dependence of subway passenger flow,it cannot effectively capture the space dependence of subway passenger flow.After converting the subway station network into an adjacency matrix,the graph convolution network can be used to deal with the spatial correlation of subway stations.By combining the graph convolutional neural network and the gated recursive unit network,an Origin-Destination Spatio-Temporal Graph Convolutional Network flow prediction model of subway stations is established,which more accurately captures the spatiotemporal correlation of subway station traffic.The convolutional network traffic prediction model of the Origin-Destination Spatio-Temporal Graph Convolutional Network is superior to the traditional model and the Long Short-term Memory Network traffic prediction model in a variety of errors,and the accuracy of the traffic prediction reaches 92.49%.This article establishes individual travel state prediction models and subway station flow prediction models to accurately predict the future urban rail traffic flow trend.The subway system manager can adjust the subway operation status based on future flow changes to alleviate the problem of possible traffic congestion.It can be seen that the study of human mobility behaviors helps to grasp the laws of human mobility,so as to better allocate urban public resources and rationally plan urban development.