Five years of transit smart card data studies

1
Five years of transit smart card data studies
Interuniversity Research Centre on Enterprise
Networks, Logistics and Transportation http//www.
cirrelt.ca

• Prof. Catherine Morency
• Prof. Martin Trépanier
• École Polytechnique de Montréal

2
In this presentation

• Background
• Studies
• Objects to analyze
• Destination estimation
• Day to day statistics
• Data mining analyses
• Comparison with household survey
• Perspectives
• Data fusion
• Activity persistency models
• Transit planning

3
Background

• Société de transport de l'Outaouais, Gatineau,
  Québec (240,000 inhab., 200 buses)
• Automated smart card fare collection system in
  place since 2001
• Contactless smart card at boarding only, on-board
  GPS for timestamp (known boarding location)
• Complete privacy of data no personal information
  available on card users
• Contactless SC now in Montreal

4
Study 1 identify the objects that can be
analyzed in the smart card information system

• Trépanier Martin, Barj S., Dufour C., Poilpré R.,
  Examen des potentialités d'analyse des données
  d'un système de paiement par carte à puce en
  transport urbain, Congrès annuel de l'Association
  des transports du Canada, Québec, Canada, 19-22
  septembre, 2004

5
S1 IS architecture
6
S1 Transportation object-oriented model
Network objects
Operations objects
Administrative objects
Demand objects
Quantities for January 2007 data
7
S1 "User" analysis
8
S1 "Network" analysis
9
S1 "Land use" analysis
10
Study 2 impute the possible destination for
each boarding transaction

• Trépanier Martin, Chapleau R., Tranchant N.,
  Individual trip destination estimation in transit
  smart card automated fare collection system,
  Journal of Intelligent Transportation Systems
  Technology, Planning, and Operations, Taylor
  Francis, Vol. 11(1), 1-15, 2007
• CHU, Ka Kee, CHAPLEAU, Robert (2008). Enriching
  archived smart card transaction data for transit
  demand modeling, Transportation Research Record
  no. 2063, pp. 63-72

11
S2 Network use analysis
12
S2 Error detection
13
S2 Vehicle block validation
14
S2 Estimation algorithm
15
S2 Estimation results
Charts? October 2003 data Success rate of 77
for 2005 data
16
Study 3 provide an intranet website for day to
day reporting on network use

• Trépanier Martin, Vassivière François,
  Democratized Smartcard Data for Transit
  Operators, 15th World Congress on Intelligent
  Transport Systems, New York, USA, 2008

17
S3 XML/SVG technologies

• XML is parsed by the web server to generate SVG
  (Scalable Vector Graphic) or HTML web pages
• SVG is displayed by the SVG plug-in
• HTML is displayed by the web browser

18
S3 Route stats
of boarding / alighting transactions
Run and departure time
Adjustment factor
Fare types
Pass-km
Maximum load
Transfers
19
S3 Load profile
filter selection
global stats on transfers and fare types
load profile for this run
boarding transactions at each stop
alighting transactions
time of passage
stop list
20
Study 4 analyze the huge quantity of smart card
data with data mining techniques

• Trépanier Martin, Morency Catherine, Agard Bruno,
  Calculation of Transit Performance Measures Using
  Smartcard Data, Journal of Public Transportation,
  (accepted)
• Morency Catherine, Trépanier Martin, Agard Bruno,
  Measuring transit use variability with smart card
  data, Transport Policy, Vol. 14(3), 193-203, 2007

21
S4 Clustering analysis
Typical workers
Early birds
Occasional users
22
S4 Group classification
WEEKS
23
S4 Network use pattern
Fare ADULT
Transit was not used
Special days 4 transactions
24
S4 Group belonging vs time
SENIOR
25
S4 Weather influence
Ridership variation per fare type
Return of spring!
ADULTS
SENIOR
STUDENTS
6 cm of snow (2.5 in.)
-24 oC(-11 oF)
7 cm of snow (3 in.)
-22 oC (-8 oF)
ice storm
26
S4 Bus stop "learning"
27
S4 Schedule adherence
2389 runs in Feb. 2005
Perfect adherence 17.5 of observations (stop
timing) 18.9 are EARLY 63.6 are LATE INBOUND
more delays
EARLY
LATE
28
S4 Network performance
Supply
Demand
29
Study 5 compare smart card data with household
survey data

• Trépanier Martin, Morency Catherine, Chapleau R.,
  Blanchette Carl, Fusion of Smart Card Data and
  Household Travel Survey Data for the Enhancement
  of Transit Riders' Behavior Characterisation, 8th
  International Conference on Survey Methods in
  Transport, Annecy, France, 2008
• Trépanier Martin, Morency Catherine, Blanchette
  Carl, Enhancing household travel surveys using
  smart card data?, 88th annual meeting of the
  Transportation Research Board, Washington, D.C.,
  2009

30
S5 two datasets
Datasets possible links between them
Size No of cars Home location Home type
Age Gender Main activity location Car
ownership Transit pass ownership
Card type Validity date
Origin Destination Purpose Departure
time Mode(s) Transit route(s)
Boarding date time Route direction Bus,
Driver Run nos Boarding stop location Alighting
stop location
Sept. 21th to Nov. 29th 2005
31
S5 Comparison by fare type
Difference between SC and HHS average day
estimation
Adult
Student
Adult express routes
Senior
HHS estimation
SC trips
32
S5 Route ridership
33
S5 Temporal distribution
34
S5 Spatial repsentation of HHS
Relative error (HHS-SC)/HHS
35
Perspectives
36
1st perspective Data fusion
relation between datasets
HHS
SC
gt
All the residents in the survey area
All public transport users
Universe
lt
App. 5 of the residing population
App. 80 of users have a smart card
Sampling rate
All transactions (boardings) on a continuous basis
lt
Observation period
Autumn period 1 autumn day by unit
All trips made by 10 years and older during 1
weekday, all transportation modes
Only boardings on public transport vehicles
(location and time)
gt
Travel indicators
Imputation of Origins and Destination points
Donor
Receptor
Variability (day to day / seasonality) of transit
use
Purpose of data fusion
Receptor
Donor
Catchment area measurements and market assessment
Donor
Donor
37
2nd perspective Activity persistency

• Longitudinal Analysis of network use
• Applying survival models to user behaviors and
  fare use
• Fare plan revenues scenarios

Example of carsharing systems
38
3rd perspective network planning

• Study of ridership variation effects (service
  adjustment)
• SC integration to demand estimation model

MADITUC 3D load profile
39
Contacts

• Catherine Morency
• cmorency_at_polymtl.ca
• Department of Civil, Geological Mines
• Martin Trépanier
• mtrepanier_at_polymtl.ca
• Department of Mathematics and Industrial
  Engineering
• École Polytechnique de Montréal, Canada