Transcript Geen diatitel

Mobile Ghent
Mobile positioning data and transport: a theoretical,
methodological and empirical discussion
24 October 2013
Bert van Wee
Delft University of Technology
The Netherlands
Presentation: focus on travel behaviour
Theoretical options follow (mainly) from data. Therefore: data
first
Not addressed, but very relevant:
• Privacy restricted versus not.
• Privacy, availability, legal aspects: probably dynamic. Role
of government very important
• Open systems: more difficult to manage
Methods / data (partly linked to theory)
•
•
•
•
•
•
•
Way more data – larger numbers, statistical significance
Cheaper
Better quality (though not always)
External quality checks
Use of ‘wisdom of the crowds’
Easier to collect
More options for (consistent) longitudinal data collection
Methods / data (continued)
•
•
•
Solution of underreporting short trips
Solution for respondents getting tired of repeatedly
reporting
Rare events / difficult to select target groups. Start selecting
people at destination (as opposed to panel / selection via
questions)
Methods (partly linked to theory)
•
Combine ‘origin based’ (persons) with destination based
(activity/destination)
•
Why would people participate? Rewards. (Airmiles)
Theory:
Why impacts theory: More data (numbers, data per person)
(non)response, disaggregation, impact behaviour
Not really fundamentally different. Nevertheless:
Theory:
• Options to test new theoretical assumptions e.g. due to
larger numbers, more data per person
• Options to discover new insights or formulate hypotheses
not based on a priori theory (Grounded Theory, data
mining). A bit risky, but also new challenges
• Options to disaggregate further (e.g. mobility trends for
specific groups of people)
Theory:
• More locational detail: enrich related theories.
• More longitudinal data: causalities.
Examples:
• Testing theory of constant Travel Time Budgets: multiple
days, also short trips. Desaggregations.
• Route choice under multiple conditions (e.g. weather)
• Mode choice in case of changing mode choice (1 person)
• Shopping behaviour (incl. fun shopping)
However
•
•
•
•
•
Practice so far: The more (bigger) data, the less theoretical
underpinnings, the less quality of analyses
Data mining
Maybe lack of awareness quality data
Ignorance of self-selection effects (e.g. leave smartphone at
home for short trips; PT: smart phone users versus others)
Privacy (may even be linked to self-selection)
Empirical
•
•
•
•
Adaptive and flexible event management
What do people do in case of emergency? Otherwise very
difficult to measure
Time space geography: action spaces: more and better data
Traffic flow (road, cars): many data, dynamics over time,
input for Satnav, short term forecasting:
1. changes in speeds, flows
2. if people would announce destination
• Walking, cycling (now often poor data)
• Travel and activities during holiday
• Better links between travel and activities:
not only ‘shopping’ but what kind of
shopping (working, recreation)
• Recreational travel behaviour (some studies
ignore recreational travel)
• Discover ‘bottlenecks’ / validate complaints
of citizens
• ‘Objective’ data for prioritization of plans
Maybe police:
• Speeding
• Drivers of lorries: too long hours?
However:
• Legal aspects
• Privacy (big brother is watching you)
Other remarks
•
•
•
We need to learn.
Risk of publication bias: only successful projects reported.
Network important!
This topic: one of many on Big (and partly open) Data.
Learn from lessons outside transport! Lot of literature in
other areas (ICT), lot of grey literature
•
•
Primary reflection: substitution for other data collection
methods. Practice: generation (new ideas, new options).
Future will show I overlooked key impacts on theory, data,
empirical options.
Reasons why we have mobile position based data has an
impact on behaviour. E.g. train instead of car because of
being online.