meili workshop: collecting travel diaries

71
MEILI: a travel diary collection system Workshop Adrian C. Prelipcean [email protected] @Adi Prelipcean adrianprelipcean.github.io 16 March 2017

Upload: adrian-c-prelipcean

Post on 14-Apr-2017

77 views

Category:

Science


7 download

TRANSCRIPT

MEILI: a travel diary collection systemWorkshop

Adrian C. Prelipcean

[email protected]

@Adi Prelipcean

adrianprelipcean.github.io

16 March 2017

Ego page

Adrian C. Prelipcean

1. Airmee - a smart platform for urban logistics– In charge of tech, R&D and innovation– Consolidate go-to market strategies with technology– Develop the Airmee ecosystem and infrastructure

2. Badger - travel and mobility data solutions– Own Badger and all MEILI products– Decide on product vision, new features and best practices– Develop and maintain the MEILI code base

3. KTH, Royal Institute of Technology– PhD student in Transportation Systems (previously GIS

[Licentiate of Tech and M. Sc.] and Geodesy [B. Eng.])– Previously TA for GIS Architecture and Algorithms, Spatial

Databases and Web&Mobile GIS– Research focused on data collection and analysis methods for

travel behaviour2

Outline

Today we will talk about:

1. Understanding travel behaviour with MEILI (Part I)– Travel behaviour and traditional collection methods– Why MEILI?– Applications of MEILI

2. Walkthrough on setting up MEILI (Part II)– Architecture of MEILI– Deployment of MEILI

I DatabaseI Travel Diary annotationI Mobility Collector

3. Best practices, security and privacy

4. Closing remarks and discussions

3

Travel behaviour

What is travel behaviour?

A vague term used by scientists when studying why, how andwhere people travel

4

Travel behaviour

How do we use travel behaviour?

Some of the main reasons for analyzing travel behaviour are:

I to investigate the reasons and mechanisms that underliean individual’s travel decision making process,

I to predict the effect of implementing new transportationpolicies or changing the transportation infrastructure, or

I to understand the dynamic of transportation movementwithin study areas.

4

Travel behaviour

How do we use travel behaviour?

Some of the main reasons for analyzing travel behaviour are:

I to investigate the reasons and mechanisms that underliean individual’s travel decision making process,

I to predict the effect of implementing new transportationpolicies or changing the transportation infrastructure, or

I to understand the dynamic of transportation movementwithin study areas.

4

Travel behaviour

How do we use travel behaviour?

Some of the main reasons for analyzing travel behaviour are:

I to investigate the reasons and mechanisms that underliean individual’s travel decision making process,

I to predict the effect of implementing new transportationpolicies or changing the transportation infrastructure, or

I to understand the dynamic of transportation movementwithin study areas.

4

(Activity) Travel diaries

What are they?

A way of summarizing where, why and how a user traveledduring a defined time frame by specifying:

I The destination of a trip

I The trip’s purposeI The means of transportation, i.e., trip legs

Img: http://soarministries.com/hp_wordpress/wp-content/uploads/2011/08/Destinations-Icon.jpg 5

(Activity) Travel diaries

What are they?

A way of summarizing where, why and how a user traveledduring a defined time frame by specifying:

I The destination of a tripI The trip’s purpose

I The means of transportation, i.e., trip legs

Img: https://cdn2.vox-cdn.com/thumbor/93Yaxs7y3Tb8tzFfppyRsSn_yN8=/1020x0/cdn0.vox-cdn.com/

uploads/chorus_asset/file/2509782/confused_man.0.jpg

5

(Activity) Travel diaries

What are they?

A way of summarizing where, why and how a user traveledduring a defined time frame by specifying:

I The destination of a tripI The trip’s purposeI The means of transportation, i.e., trip legs

Img: https://d3ui957tjb5bqd.cloudfront.net/images/screenshots/products/4/42/42990/

white-transportation-icons-300x200.jpg

5

Collecting travel diaries

Traditionally

I Users declare what they have done in a travel survey, e.g.,PP or CATI

I Issues with traditional methods:– Expensive to organize large data collection sessions– It is difficult to centralize human-readable free-form data– Systematically biased towards older age groups– Decreasing response rates– Depends on respondents’ memory regarding the day, which

mostly affects short trips– Difficult to notice trends due to very short data collection

horizon (usually one day)– Limited amount of information (e.g., the only information on

travel modes is the main travel mode, no information onwaiting times, etc.)

– Filling in fatigue

Img: http://www.schoolsurveyexperts.co.uk/i/photos/paper_survey.jpg

6

Collecting travel diaries

Traditionally

I Users declare what they have done in a travel survey, e.g.,PP or CATI

I Issues with traditional methods:– Expensive to organize large data collection sessions– It is difficult to centralize human-readable free-form data– Systematically biased towards older age groups– Decreasing response rates– Depends on respondents’ memory regarding the day, which

mostly affects short trips– Difficult to notice trends due to very short data collection

horizon (usually one day)– Limited amount of information (e.g., the only information on

travel modes is the main travel mode, no information onwaiting times, etc.)

– Filling in fatigue

6

Collecting Travel diaries

A different way to collect them

I MEILI - GPS collection + Web and Mobile GIS basedinteraction + Artificial Intelligence / Machine Learning

7

MEILI - a system for collecting travel diaries

What problems is MEILI soving?

I Allows for the seamless collection of a large number ofusers for a longer duration of time

I Data are already centralized during the collection stage

I Travelers are asked to validate data already annotated byMEILI, which minimizes fatigue and the possibility offorgetting about a trip

I Displays collected data to travelers

I Very detailed granularity for the collected data (routes,triplegs, waiting times, etc.)

I It is of considerably lower cost to organize longer datacollection sessions

8

MEILI - a system for collecting travel diaries

What problems is MEILI not soving?

I To decrease response rates, one should investigate howincentives can be paired with MEILI.

I Age group bias is still a problem, since annotating datavia a website needs some technical knowledge that is notgranted for all age groups

9

MEILI - a system for collecting travel diaries

10

MEILI - a system for collecting travel diaries

11

Data collected by MEILI

12

Overview of MEILI case studies

MEILI is used in different case studies for different purposes,such as:

I comparing semi-automated travel diary collection systemswith traditional collection methods in Stockholm, Sweden(3 case studies in 2013, 2014 and 2015)

I testing the ease of transferability of MEILI toGothenburg, Sweden (1 case study in 2016)

I monitoring active mobility for school children inSingapore (currently trialled for 1 case study)

I tracking delivery patterns in Singapore (currently trialledfor 1 case study)

13

Stockholm - Comparing MEILI with traditional

collection methods (PP)

I Partners: KTH, Sweco, LinkopingUniversity

I Number of participants: 11 (first casestudy), 30 (second case study), 171(third case study)

14

Stockholm - Comparing MEILI with traditional

collection methods (PP)

I Partners: KTH, Sweco, LinkopingUniversity

I Number of participants: 11 (first casestudy), 30 (second case study), 171(third case study)

I The overall descriptive statistics forboth systems are similar.

14

Stockholm - Comparing MEILI with traditional

collection methods (PP)

I Partners: KTH, Sweco, LinkopingUniversity

I Number of participants: 11 (first casestudy), 30 (second case study), 171(third case study)

I The overall descriptive statistics forboth systems are similar.

I The percentage of trips captured byMEILI increased between the two casestudies.

14

Stockholm - Comparing MEILI with traditional

collection methods (PP)

I Partners: KTH, Sweco, LinkopingUniversity

I Number of participants: 11 (first casestudy), 30 (second case study), 171(third case study)

I The overall descriptive statistics forboth systems are similar.

I The percentage of trips captured byMEILI increased between the two casestudies.

I The reasons for missing a tripchanged between the two case studies.

14

Stockholm - Comparing MEILI with traditional

collection methods (PP)

I MEILI collects more trips and at afiner granularity, but it does notcollect all trips

I There is no clear superior method forcollecting travel diaries

See:

I Allstrom et al. (2016) ”Experiences fromsmartphone based travel data collection -System development and evaluation”

I Susilo et al. (2016) ”Lessons from a trial ofMEILI, a smartphone based semi-automaticactivity-travel diary collector, in Stockholmcity, Sweden” in Proceedings of WCTR 2016.

I Prelipcean et al. (2017). ”A series of threecase studies on the semi-automation ofactivity travel diary generation usingsmarpthones” in Proceedings of TRB 2017Annual Meeting. 15

Overview of applying MEILI to solve research

problems

Using the data collected with MEILI, different research issueswere investigated, such as:

I Developing methods to compare different travel diarycollection systems

I Interdisciplinary travel behaviour analysis

I Automating the generation of travel diaries from GPStrajectories

I Making use of multi-day data to extract behaviouralpatterns

16

Comparing different travel diary collection systems

I Trips best matched on time andpurpose - geocoding is error prone

I Types of captured information:– Intrinsic - how well an entity, incl. its

attributes, is captured?– Extrinsic - how well do different

systems agree on the capture of anentity?

I Spatial and temporal indicatorsmeasure intrinsic information

I Intrinsic and extrinsic informationuseful for in-depth analysis

I Unifying framework of previousconcepts

17

Comparing different travel diary collection systems

I Trips best matched on time andpurpose - geocoding is error prone

I Types of captured information:– Intrinsic - how well an entity, incl. its

attributes, is captured?– Extrinsic - how well do different

systems agree on the capture of anentity?

I Spatial and temporal indicatorsmeasure intrinsic information

I Intrinsic and extrinsic informationuseful for in-depth analysis

I Unifying framework of previousconcepts

17

Comparing different travel diary collection systems

I Trips best matched on time andpurpose - geocoding is error prone

I Types of captured information:– Intrinsic - how well an entity, incl. its

attributes, is captured?– Extrinsic - how well do different

systems agree on the capture of anentity?

I Spatial and temporal indicatorsmeasure intrinsic information

I Intrinsic and extrinsic informationuseful for in-depth analysis

I Unifying framework of previousconcepts

17

Comparing different travel diary collection systems

I Trips best matched on time andpurpose - geocoding is error prone

I Types of captured information:– Intrinsic - how well an entity, incl. its

attributes, is captured?– Extrinsic - how well do different

systems agree on the capture of anentity?

I Spatial and temporal indicatorsmeasure intrinsic information

I Intrinsic and extrinsic informationuseful for in-depth analysis

I Unifying framework of previousconcepts

17

Comparing different travel diary collection systems

I Trips best matched on time andpurpose - geocoding is error prone

I Types of captured information:– Intrinsic - how well an entity, incl. its

attributes, is captured?– Extrinsic - how well do different

systems agree on the capture of anentity?

I Spatial and temporal indicatorsmeasure intrinsic information

I Intrinsic and extrinsic informationuseful for in-depth analysis

I Unifying framework of previousconcepts

17

Comparing different travel diary collection systems

I Trips best matched on time andpurpose - geocoding is error prone

I Types of captured information:– Intrinsic - how well an entity, incl. its

attributes, is captured?– Extrinsic - how well do different

systems agree on the capture of anentity?

I Spatial and temporal indicatorsmeasure intrinsic information

I Intrinsic and extrinsic informationuseful for in-depth analysis

I Unifying framework of previousconcepts

See: Prelipcean et al. (2015). ”Comparativeframework for activity-travel diary collectionsystems”, in Proceedings of MT-ITS 2015, pages251-258, DOI: 10.1109/MTITS.2015.7223264.

17

Different ways of looking at travel

Different fields that have different views ontravel:

I Transport Science - Q: How were userstravelling during a defined period?

18

Different ways of looking at travel

Different fields that have different views ontravel:

I Transport Science - Q: How were userstravelling during a defined period?

I Location Based Services - Q: How is auser travelling now?

18

Different ways of looking at travel

Different fields that have different views ontravel:

I Transport Science - Q: How were userstravelling during a defined period?

I Location Based Services - Q: How is auser travelling now?

I Human Geography - Q: How can atrajectory be segmented into partsthat can be enriched with domainspecific semantics?

18

Different ways of looking at travel

Different fields that have different views ontravel:

I Transport Science - Q: How were userstravelling during a defined period?

I Location Based Services - Q: How is auser travelling now?

I Human Geography - Q: How can atrajectory be segmented into partsthat can be enriched with domainspecific semantics?

I Every domain has a unique andnon-transferable definition of error

See: Prelipcean et al (2016). ”Transportation modedetection – an in-depth review of applicability andreliability”, in the Journal of Transport Reviews,ahead of print, 2016, DOI:10.1080/01441647.2016.1246489.

18

Robust errors for trajectory segmentations

Efforts towards obtaining robust errormeasures for trajectory segmentation:

I Replaced rigid interval matching withpenalty-based interval alignment

I Proposed three mode detectionstrategies:

– Implicit segmentation– Explicit-holistic segmentation– Explicit-consensus based segmentation

I Introduced new travel modesegmentation performance metrics:

– Precision and Recall– Shift-in and Shift-out penalties– Oversegmentation

19

Robust errors for trajectory segmentations

Efforts towards obtaining robust errormeasures for trajectory segmentation:

I Replaced rigid interval matching withpenalty-based interval alignment

I Proposed three mode detectionstrategies:

– Implicit segmentation– Explicit-holistic segmentation– Explicit-consensus based segmentation

I Introduced new travel modesegmentation performance metrics:

– Precision and Recall– Shift-in and Shift-out penalties– Oversegmentation

19

Robust errors for trajectory segmentations

Efforts towards obtaining robust errormeasures for trajectory segmentation:

I Replaced rigid interval matching withpenalty-based interval alignment

I Proposed three mode detectionstrategies:

– Implicit segmentation– Explicit-holistic segmentation– Explicit-consensus based segmentation

I Introduced new travel modesegmentation performance metrics:

– Precision and Recall– Shift-in and Shift-out penalties– Oversegmentation

19

Robust errors for trajectory segmentations

Efforts towards obtaining robust errormeasures for trajectory segmentation:

I Replaced rigid interval matching withpenalty-based interval alignment

I Proposed three mode detectionstrategies:

– Implicit segmentation– Explicit-holistic segmentation– Explicit-consensus based segmentation

I Introduced new travel modesegmentation performance metrics:

– Precision and Recall– Shift-in and Shift-out penalties– Oversegmentation

See: Prelipcean et al (2016). ”Measures oftransport mode segmentation of trajectories”, in theInternational Journal of Geographical InformationScience, Volume 30, Issue 9, pages 1763-1784,2016, DOI: 10.1080/13658816.2015.1137297. 19

Travel diary generation from trajectories

Automating the travel diarygeneration from trajectories:

I Destination inference performanceinsufficient because of the POIsassociated with user activities

I Purpose inference capped bydestination inference

I Travel mode segmentation

20

Travel diary generation from trajectories

Automating the travel diarygeneration from trajectories:

I Destination inference performanceinsufficient because of the POIsassociated with user activities

I Purpose inference capped bydestination inference

I Travel mode segmentation

20

Travel diary generation from trajectories

Automating the travel diarygeneration from trajectories:

I Destination inference performanceinsufficient because of the POIsassociated with user activities

I Purpose inference capped bydestination inference

I Travel mode segmentation difficultwith more travel modes:

– Good at a user level (personallearning)

– Learning from the population historyhas minor gains

– Tripleg detection caps modesegmentation inference

20

Travel diary generation from trajectories

Automating the travel diarygeneration from trajectories:

I Destination inference performanceinsufficient because of the POIsassociated with user activities

I Purpose inference capped bydestination inference

I Travel mode segmentation:– Good at a user level (personal

learning)– Learning from the population history

has minor gains– Tripleg detection caps mode

segmentation inference

See: Prelipcean(2016) ”Capturing travel entities tofacilitate travel behavior analysis - A case study ongenerating travel diaries from trajectories”,Licentiate Thesis in Geoinformatics andTransportation Science.

20

Sequential stability of travel behaviour activities

I It is intuitive to notice simple schedulepatterns

I We can extract the Longest CommonSubsequences to identify commonpatterns between days and travelers

I Half of activities are performed in thesame order in the user base

I The travel mode scheduling is morediverse, especially on weekends

I High similarity for travel modes whenperforming the same activity, but onlyon the intra-personal level

21

Sequential stability of travel behaviour activities

I It is intuitive to notice simple schedulepatterns

I We can extract the Longest CommonSubsequences to identify commonpatterns between days and travelers

I Half of activities are performed in thesame order in the user base

I The travel mode scheduling is morediverse, especially on weekends

I High similarity for travel modes whenperforming the same activity, but onlyon the intra-personal level

21

Sequential stability of travel behaviour activities

I It is intuitive to notice simple schedulepatterns but the complexity increaseswith the number of days and activities

I We can extract the Longest CommonSubsequences to identify commonpatterns between days and travelers

I Half of activities are performed in thesame order in the user base

I The travel mode scheduling is morediverse, especially on weekends

I High similarity for travel modes whenperforming the same activity, but onlyon the intra-personal level

21

Sequential stability of travel behaviour activities

I It is intuitive to notice simple schedulepatterns but the complexity increaseswith the number of days and activities

I We can extract the Longest CommonSubsequences to identify commonpatterns between days and travelers

I Half of activities are performed in thesame order in the user base

I The travel mode scheduling is morediverse, especially on weekends

I High similarity for travel modes whenperforming the same activity, but onlyon the intra-personal level

21

Sequential stability of travel behaviour activities

I It is intuitive to notice simple schedulepatterns but the complexity increaseswith the number of days and activities

I We can extract the Longest CommonSubsequences to identify commonpatterns between days and travelers

I Half of activities are performed in thesame order in the user base

I The travel mode scheduling is morediverse, especially on weekends

I High similarity for travel modes whenperforming the same activity, but onlyon the intra-personal level

21

Sequential stability of travel behaviour activities

I It is intuitive to notice simple schedulepatterns but the complexity increaseswith the number of days and activities

I We can extract the Longest CommonSubsequences to identify commonpatterns between days and travelers

I Half of activities are performed in thesame order in the user base

I The travel mode scheduling is morediverse, especially on weekends

I High similarity for travel modes whenperforming the same activity, but onlyon the intra-personal level

21

Sequential stability of travel behaviour activities

I It is intuitive to notice simple schedulepatterns but the complexity increaseswith the number of days and activities

I We can extract the Longest CommonSubsequences to identify commonpatterns between days and travelers

I Half of activities are performed in thesame order in the user base

I The travel mode scheduling is morediverse, especially on weekends

I High similarity for travel modes whenperforming the same activity, but onlyon the intra-personal level

See: Prelipcean et al (2017). Longest commonsubsequences: Identifying the stability ofindividuals’ patterns. Working paper.

21

Part II - Walkthrough

Overview of MEILI Architecture

23

MEILI Database

Roles:

I Stores the raw GPS and accelerometer data collected bythe mobile phones (via Mobility Collector)

I Stores the annotations for trips and triplegs

I Enables the CRUD operations of the API

Minimum Req: Postgres 9.3 [link] and PostGIS 2.1.7 [link].

Extra requirements:

I POI dataset for destinations (OSM)

I POI dataset for transportation stops and parking places(OSM)

Github repo: https://github.com/Badger-MEILI/MEILI-DatabaseSee: Prelipcean et al. (2017). ”MEILI: an activity travel diary collection, annotationand automation system” submitted to Journal of Urban Technology

24

MEILI Database Setup

Deployment:

1. Create empty database for MEILI

$ createdb meili_demo_db

2. Clone the MEILI database repository$git clone [email protected]:Badger-MEILI/MEILI-Database.git

3. Remove git details$rm .git -R

4. Initialize the MEILI database with the functions in the initscript$ psql -U user_name -d meili_demo_db -a

-f SQL/init.sql -v ON_ERROR_STOP=1

5. Run tests with pgTAP [link]$ cd Unit_Tests && ./script.sh meili_demo_db user_name

25

MEILI Database Hosting

Possible hosting solutions

I Own server

I Amazon Postgres RDS [link]

I Amazon EC2 instance with custom Posgres installation[link]

I Openshift online with Postgres cartridge [link]

I Heroku with Postgres addon [link]

I Other providers

Pay extra attention to:

I encryption

I connection pooling

26

MEILI Travel Diary and API

Roles:

I Users visualize their data

I Endpoints for data upload

I Users correct wrong annotations

I Good place to implement extensions for user retention

Built with the NodeJs (v6.10.0 LTS) [link]

Extra requirements:

I SSL on the hosting server

I Make sure IP and port are accessible

Github repo: https://github.com/Badger-MEILI/MEILI-Travel-Diary

See: Prelipcean et al. (2017). ”MEILI: an activity travel diary collection, annotation

and automation system” submitted to Journal of Urban Technology

27

MEILI Travel Diary and API Setup

1. Clone the MEILI Travel Diary repository$git clone [email protected]:Badger-MEILI/MEILI-Travel-Diary.git

2. Remove git details$rm .git -R

3. Install Node packages$ npm install

4. Add your own database credentials:$ vim routes/database.js

5. Start Travel Diary Web app$ npm start

6. Open up another terminal and run tests$ npm run test-client

28

MEILI Travel Diary and API Hosting

Hosting:

I Own server

I Amazon EC2 with NodeJS installed [link]

I Openshift with NodeJS cartridge [link]

I Heroku with NodeJS [link]

I Other providers...

Pay extra attention to:

I secured communication (HTTPS or HTTP 2.0)

I updating Node packages for security and stability

29

MEILI Mobility Collector

Roles:

I Allows users to register

I Collects GPS and accelerometer data in a battery efficientway (battery consumption related to the time spenttraveling)

I Periodically sends GPS and accelerometer data to theserver

Built on: Android (min tested version Android 5.0 Eclair) andiOS (min tested version iOS 8)

Extra requirements:

I Android - build tools in Android SDK (any platform) [link]

I iOS - XCode for iOS (Mac OS only) [link]

30

MEILI Mobility Collector Deployment

Official distribution:I Google Play Store [link]

– Android only– app is ready within a day (if there are no copyright issues)– can be distributed to any number of users– Invitation via official URL

I Apple App Store [link]– iOS only– 50% of apps are reviewed in 24 hours and 90% are reviewed in

48 hours after submission and accepted if compliant– can be distributed to any number of users (if accepted)– Invitation via official URL

31

MEILI Mobility Collector Deployment

Test distribution:I Beta by Crashlytics (Fabric and Google) [link]

– iOS and Android– available immediately on Android, unlimited number of users– available immediately on iOS, complicated invite process for

iOS, limited to 100 users– Invitation via email or onboarding URL

I Google Testing in Play store [link]– similar to official distribution, only you choose who to invite– available immediately

I Apple TestFlight [link]– iOS only– app is reviewed within the day (usually), limited to 2000 users– invitation via email (linked to Apple ID)

Github repo: https://github.com/Badger-MEILI/MEILI-Mobility-Collector-iOS andhttps://github.com/Badger-MEILI/MEILI-Mobility-Collector-Android

See: Prelipcean et al. (2014). ”Mobility Collector”, in the Journal of Location BasedServices, Volume 8, Issue 4, pages 229-255, DOI: 10.1080/17489725.2014.973917.

32

MEILI best practices

Minimize your worries :I Alfa stage

1. Get at least one technical person with good overall knowledgeof development and operations in the project

2. Validate your collection strategy on a small user baseconsisting of technical people close to the project

3. Validate the collection on the device types you are targeting4. Identify points of failure5. Go through each point of failure, decide on a recovery

strategy, make sure it works and document the strategy6. Perform a simplified version of the analysis you want to

perform7. Decide on the data that are meaningful to log (latency,

connectivity, API calls, errors, etc.)

33

MEILI best practices

Minimize your worries :

I Alfa stageI Beta stage

1. Organize a pilot test, aim to have around 10% of the numberof users targeted in the case study, with users that can accepta product that continuously evolves

2. Test and adapt the defined recovery strategies to the newusers

3. Analyze the learning curve of new users: time to firstannotation and time to first inference are important here

4. Ideally, run the beta study for the same duration of time youintend to run the main study for

5. Design alarms for errors and critical values for what youdecided to log

33

MEILI best practices

Minimize your worries :

I Alfa stage

I Beta stageI Case study stage

1. Have the recovery strategies with concrete examples of how toapply them in an easily accessible place for everyone in theteam

2. Have user support ready3. Have technical personnel on call4. If possible, gradually invite your users in 10% batches

33

Security and privacyI Do not collect data you are not using (no nice to have) -

you are liable for the data you collect

I Explain why you collect data and how you plan onanalyzing it in a easy to read disclaimer

I Have dedicated support for users to contact you for anydata inquiry

I Be compliant with the national and international privacylaws

I Do not assume expertise, contact security and privacyexperts before starting a project

34

Problems with using MEILII You do not benefit from around-the-clock support

I You need to provide your own support for your users

I No solution for localization yet

I Collection capabilities can be disrupted by new OSversions for the smartphones you are targeting

I You will need a dedicated developer with ops knowledgefor the duration of your case studies

I Hosting costs, although most hosting solutions have afree offering that can support MEILI for hundreds of users

I You have to decide on which POI dataset to use andmake it compliant to the existing database form

I No socio-demographics data collected by default

I Users have complained about UI and UX

35

Benefits of using MEILII There is no development cost to get up and running

I You own the data you collect

I You benefit from the shared development efforts due toopen-source licensing as copyleft

I You can invest in extending MEILI as you see fit (e.g.,collect socio-demographics, embed survey questions, etc.)

I You can be compliant to any security and privacyregulations as you have full control of the source code

I You can modify MEILI Mobility Collector to use othersmartphone sensors (e.g., temperature, pressure,illumination, etc.)

I You can make it your own way

36

Thank you for your attention!Questions and Discussions

Adrian C. Prelipceanhttp://adrianprelipcean.github.io/[email protected]@Adi Prelipcean

This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 3.0 Unported License.

References

References (part I)I source code for the MEILI family

– Database: https://github.com/Badger-MEILI/MEILI-Database

– Travel Diary and API:https://github.com/Badger-MEILI/MEILI-Travel-Diary

– Mobility Collector for Android https:

//github.com/Badger-MEILI/MEILI-Mobility-Collector-Android

– Mobility Collector for iOShttps://github.com/Badger-MEILI/MEILI-Mobility-Collector-iOS

I papers on MEILI architecture and data collection– A. C. Prelipcean, G. Gidofalvi, and Y. Susilo. 2014. ”Mobility Collector”,

in the Journal of Location Based Services, Volume 8, Issue 4, pages229-255, DOI: 10.1080/17489725.2014.973917

– A. C. Prelipcean, G. Gidofalvi, and Y. Susilo. ”MEILI: an activity traveldiary collection, annotation and automation system” submitted toJournal of Urban Technology

38

References

References (part II)I papers and technical projects on MEILI case studies

– A. C. Prelipcean, G. Gidofalvi, and Y. Susilo. 2017. ”A series of threecase studies on the semi-automation of activity travel diary generationusing smarpthones” in Proceedings of TRB 2017 Annual Meeting, aheadof print. [link]

– A. Allstrom, G. Gidofalvi, I. Kristoffersson, A. C. Prelipcean, C.Rydergren, Y. Susilo, J. Widell. 2016. ”Experiences from smartphonebased travel data collection - System development and evaluation”, Finalreport for the SPOT-project. [link]

– A. Allstrom, A. C. Prelipcean, M. Gejdeback, T. Skoglund. 2016.”Erfarenheter fran forsok med smartphone-baserad resdatainsamling iGoteborg”, Final report for the SPOT Gothenburg-project.

– A. C. Prelipcean. 2016. ”Capturing travel entities to facilitate travelbehavior analysis - A case study on generating travel diaries fromtrajectories” Licentiate Thesis in Geoinformatics and TransportationScience. [link]

– Y. Susilo, A. C. Prelipcean, A. Allstrom, G. Gidofalvi, I. Kristoffersson, J.Widell. 2016. ”Lessons from a trial of MEILI, a smartphone basedsemi-automatic activity-travel diary collector, in Stockholm city,Sweden.”, Proceedings of WCTR 2016.

39

References

References (part III)I papers on data analysis based on MEILI collected data

– Comparing travel diary collection systemsI A. C. Prelipcean, G. Gidofalvi, and Y. Susilo. 2015. ”Comparative

framework for activity-travel diary collection systems”, inProceedings of MT-ITS 2015, pages 251-258, DOI:10.1109/MTITS.2015.7223264. [link]

– Robust trajectory segmentation errorsI A. C. Prelipcean, G. Gidofalvi, and Y. Susilo. 2016. ”Measures of

transport mode segmentation of trajectories”, in the InternationalJournal of Geographical Information Science, Volume 30, Issue 9,pages 1763-1784, DOI: 10.1080/13658816.2015.1137297. [link]

– Sequential stability of travel behaviourI A. C. Prelipcean, Y. Susilo, G. Gidofalvi. Longest common

subsequences: Identifying the stability of individuals’ patterns.Working paper.

– Interdisciplinary literature review on travel mode detectionI A. C. Prelipcean, G. Gidofalvi, and Y. Susilo. 2016.

”Transportation mode detection – an in-depth review ofapplicability and reliability”, in the Journal of Transport Reviews,ahead of print, DOI: 10.1080/01441647.2016.1246489. [link]

40