Big Data for Travel Demand Modeling

DataPalooza 6/4/2014 : Analyzing Integrated Data Part 2

Transcript

1. Dr. Laurie A. Schintler The School of Public Policy George Mason University Draft
2. What is Big Data? Big Data for Travel Demand Modeling • Opportunities • Strengths • Examples • Weaknesses • Threats Conclusions and Discussion
5. Source: marketingpilgrim.com Putting a Petabyte into Perspective
7. USE OF DIGITAL SERVICES mobile phones Web services ONLINE INFORMATION social media newspapers, e-commerce transactions PHYSICAL SENSORS satellite imagery nightlights data CROWDSOURCING Volunteered Geographic Information Structured Versus Unstructured Data Varying Levels of spatial and temporal variation Repurposed Data Sources Data Characteristics
8. … VELOCITY
9. • Noise • Incompleteness • Biased • Anomalies/outliers • Uncertainty VERACITY Trustworthiness of the data Wikipedia.com Apollo.lsc.vsc.edu Baldscientist.wordpress.com
11. Mashable.com Nist.com Wikipedia.com
12.  Human/Societal Behavior  Social Networks  Proliferation of Unstructured Data (text, video, photos)  Digitization of documents and content  Sentiment/Preferences  Crowdsourcing What’s New? Ripe for Travel Demand Analysis (And Transportation In General) What’s It For Us?
13. 0 9 29 172 0 50 100 150 200 2011 2012 2013 2014 TRB Papers Associated with "Big Data" Source: Shanjiang Zhu, George Mason University
14. Innovations In Travel Modeling, Baltimore, 04/27-04/30 Innovations in Data Workshop Dynamic Models and Dynamic Data Old Data with a New Twist GPS and Emerging Data in Travel Forecasting International Transport Forum 2014 Summit, Leipzig •Big Data in Transport: Applications, Implications, Limitations Exploratory Advanced Research •Video feature extraction automation •Real time data for connected highway-vehicle systems •FY 2014 Solicitation, Data for safety analysis and freight Source: Shanjiang Zhu, George Mason University …And of course Transportation Datapalooza!
16. What are we trying to understand? …And why? • When and why do people travel? • What mode of transport do they use? • What routes do they take? • Do they change their travel behavior under certain circumstances? • Do they travel with others? • How do activities drive travel behavior? • How does travel behavior vary depending on characteristics of individuals? • What are the preferences of travelers? • Capacity Planning/Expansions • Transportation Demand Management • Strategies to Enhance Accessibility (Including Disadvantaged Segments of the population) • Livable/Sustainable Solutions • Democratization of Transport Planning/Policy Process
17. www.aimsun.com • Household Surveys • Travel Diaries • Loop detectors • Census Data
18. TRADITIONAL DATA BIG DATA  Limited geographic and temporal coverage  Lack of consistency across regions  Lack information on socialization/social ties  Small samples, margin of error  Short-Distance Travel  Information on Residents  Costly, Timely to Collect  Very fine-grained levels of resolution  Available for many, if not all parts of the globe  Social networks  Large Datasets  Long-Distance  Information on Visitors (In addition to Residents)  Costly, Timely to Process
20. Mobile Phone Traces Location-Sharing Services (LSS) data Taxi Movements/Public Transportation Targeted Market Data
21. Sources: c2.staticflickr.com More refined locations can be defined using information on: • Speed of Travel • Stops • Development Patterns
22. From: Cheng et al. (2011) Foursquare Checkins (October 2010 – February 2011) Foursquare, Gowalla, Brightkite Data: “checkins”, social networks
23. Wikipedia.com Dataconomy.com Wikipedia.com
24. Phx.homeguide.com • Internet clicks/searche s • Use of mobile apps • Consumer Purchases • Text Messages Collecting Massive Amounts of Data About Us!
25. Atlanta Chicago Source: Schintler et al., 2011
26. Flowingcity.com Taxi Movements in London Taxis in NYC Source: Ferraira, Poco, Freire, Silva (2013), Visual Exploration of Big Spatio-Temporal Urban Data: A Study of NY Taxi Trips
27. Source: Gonzalez M, Hidalgo C, Barabasi AL (2008) Understanding individual human mobility patterns. Nature Letters. 5: 779-782 Cell Phone Traces Within a Region
28.  Anonymous Cell Phone Records and Mobile Phone traces, “Human mobility and resulting contacts between people have a high degree of predictability (Song et al., 2010)”  Regular patterns across population, despite the different socio-economic characteristics Of Individuals and travel context  Scaling in distances travelled, time spent at locations and popularity of places  Artefact of Underlying Development or Do We Care? Example of Scaling Behavior Source: brenocom.com
29. When People Travel Within a Day How Far People Travel Source: Cheng et al. (2011), Exploring Millions of Footprints in Location Sharing Services, Proceedings of the Fifth International AAAI Conference on Weblogs and Social Media
30. “Socio-Spatio” Network approach 1 set of vertices represent space, the other set represent individuals The full dataset contains 4,491,143 “check-ins” made by Brightkite users over the period April 2008-October 2010. The undirected social graph is comprised of 58,228 individuals and 214, 078 social ties (Cho et al. 2011). We extracted “check-ins for only t hose locations in the contiguous United States, and assigned “check-ins” to counties. There were almost 3 million “check- ins, 21,001 unique users and 2706 counties with at least one “check-in.” Schintler, L.A., Kulkarni, R. Haynes, K. and R. Stough (2014). Sensing “Socio-Spatio” Interaction and Accessibility from Location-Sharing Services Data. eds. Margerida, A. et al. Accessibility and Spatial Interaction
31.  How does spatial proximity and travel behavior affect probability of friendship ?  How do social ties influence travel behavior?
32. Josie Kressner, Georgia Tech Targeted Marketing Data Synthetic Data – Capture All Aspects of Travel Behavior
34.  Not random: bias  Missing data (and noise!)  Observations not independent  Issues related to temporal/spatial data  Traditional Methods Do Not Apply  Inequities  Computational Complexity www.thesalesblog.com
35. Chen and Schintler, forthcoming More research is needed to assess the statistical accuracy of big data, what types of bias are present, how can they be corrected for, and what analytical methods are best-suited for the data!!
37. • How far can we go in the use of big data for understanding/modeling travel behavior? • How do we safeguard individual’s privacy? www.publicpolicy.telfonica.co m www.cbsnews.com