Spatio-temporal Travel Pattern Mining And Dynamic Passenger Flow Analysis In Urban Rail Transit System

In recent years,urban transportation is facing with the problems of traffic safety,traffic congestion and energy pollution.Giving priority to the develop-ment of public transportation is the important measure to solve the problem,and developing the urban rail transit is the important aspect.With the rapid development of urban rail transit system,it is becoming more complex on the aspects of rail network structure,traffic demand,space-time distribution of pas-sengers.However,accurate knowledge of travel pattern and dynamic distribution of passenger flow are the premise for transit resource allocation,emergency co-ordination,strategy formulation and evaluation.With extensive use of AFC(automated fare collection)for urban transit sys-tem and the accessibility of other related data(such as train schedule,mete-orological data),the data that we could obtain has qualitative improvement in terms of quality,size,type,timeliness and other aspects.How to extract valuable knowledge from these massive data has become an important research topic in the field of spatio-temporal data management and mining.This paper analyses urban rail traffic from four aspects:passenger spatio-temporal travel pattern analysis,passenger route choice behavior analysis,dynamic OD(origin-destination)matrix estimation,dynamic passenger flow movement analysis.(1)In-depth analysis and mining on passengers' spatio-temporal travel pat-terns:based a set of massive smart card data over a long period,we study individual's general travel style and regularity travel patterns in terms of space and time.Based on three aspects,temporal,spatial and spatio-temporal,we de-fine individual's travel patterns and we propose methods to retrieve the pattern-s.Then,for anomaly detection and regularity discovery,we analyze individual passengers' travel patterns by statistical-based and unsupervised cluster-based methods.From statistical-based point of view,we look into the passenger travel distribution patterns and find out the abnormal passengers based on the empirical knowledge.From unsupervised clustering point of view,we classify passengers in terms of the similarity of their travel patterns.To interpret the group behaviors,we also employ the bus transaction data.Moreover,the abnormal passengers are detected based on the clustering results.We give some reasonable explanations for these abnormal passengers through surveys.(2)Passenger route choice behavior analysis:In this paper,we propose a solution that needs no additional facility than the trains operating time table and the AFC records data to calculate the probability of each route chosen for an OD pair.We first calculate two kinds of time-dependent polynomial distributions of the number of trains waited for by passengers.We further propose a probabilistic model that can estimate how the passenger flows are distributed among different routes and trains.(3)Dynamic passenger flow OD matrix estimation:Considering the closed network structure of urban rail system,the high reliability of travel time,the deterministic and stochastic characteristics of passenger travel patterns,we use inflow,outflow and weather as real-time detection information to estimate dy-namic OD matrix.We first extract spatio-temporal travel patterns of each in-dividual passenger and classify passengers based on whether the destination can be inferred or not into fixed passengers and stochastic passengers.For predict-ing the destination of stochastic passengers,a hybrid WAM(Weighted average method)/KNN(K-Nearest Neighbor)model was introduced.It handles the lin-ear and nonlinear characteristic of passenger flow using non-overlapped features,then give different weighting coefficient to WAM and KNN according to cross validation.(4)Dynamic passenger flow movement analysis:Based on AFC records,trains schedule,meteorological data,we construct an online passenger movement analysis system.We propose a state transition model to estimates passengers'movement inside rail system.The influence factors of state transition are an-alyzed,such as distribution of passenger's walking time;the line a passenger chooses after entering a station with multiple lines;the probability for on-train passengers transiting to other states when a train reaches a station.