| As an important group of modern urban residents,public transport passengers can grasp the travel situation of public transport passengers in advance,which can not only enable decision-makers to make dynamic adjustments in time according to the changes of passenger travel demand in transportation planning,but also provide decision-making basis for targeted vehicle advertising,infectious disease prevention and urban planning.The research object of this paper is the travel space-time behavior of urban bus passengers.The purpose of this study is to apply deep learning to predict the future travel space-time behavior of bus passengers according to the historical travel rules of bus passengers and bus travel characteristics.A complete prediction method of bus passenger travel space-time behavior is proposed.In this paper,the spatio-temporal behavior of bus passengers is expressed through the bus trip-chain,and the prediction idea of the future trip-chain of bus passengers is introduced in detail.The prediction of bus trip-chain is divided into six sub-questions: whether passengers will travel,the number of trips,the boarding station,the boarding time,the alighting station and the alighting time.According to the principle of deep learning model,the analysis process of each sub-problem is parameterized.The corresponding input feature set is selected according to the sub-problem,and the artificial neural network(Artificial Neural Network,ANN)is used to construct the prediction model of future travel behavior of bus passengers.In this paper,using the Pytorch deep learning open source framework and Python programming language as the development environment,using the intelligent bus(Advanced Public Transportation Systems,APTS)data,(Point Of Interest,POI)distribution data and historical weather records for 2 weeks in Weinan,China,the fully connected prediction model is trained and verified,the goal of predicting the future travel space-time behavior of urban bus passengers is achieved,and the specific impact of each bus travel characteristics on the final travel forecast results is mastered.Through theoretical research and case comparison,it is found that the prediction method of bus passenger travel spatio-temporal behavior proposed in this paper is effective and practical,and can be applied to other similar studies. |
