Page 67 - Proceeding of Atrans Young Researcher's Forum 2019_Neat

“Transportation for A Better Life:
Smart Mobility for Now and Then”

23 August 2019, Bangkok, Thailand

5. Conclusion 6. Acknowledgment
This paper presented preliminary results of The authors highly appreciate anonymous
a smartphone-based mobility survey experiment reviewers’ constructive comments that help to
conducted by DEST in Hanoi, Vietnam recently. strengthen the final version. And they would like to
Participants have welcomed and participated give a big thank to organizers of ATRANS annual
actively in the survey. Issues related to technology, conference that is such an interesting and useful
battery and privacy concern are in line with previous environment of young researchers.
findings summarized in [3]. The typical challenges
in terms of validation and language for case of Hanoi References
stem from using the app designed for European [1] J. Armoogum et al., Survey Harmonisation with
countries. However, their adverse effects were New Technologies Improvement (SHANTI).
limited thanks to the kind and timely supports from IFSTTAR, 2014.
DEST. This experiment contributed the [2] J. Armoogum, K. W. Axhausen, and J.-L. Madre,
geographical diversity of smartphone-assisted “Chapter 36 Lessons from an Overview of National
mobility survey field because most of previous ones Transport Surveys, from Working Group 3 of COST
were conducted in developed countries. 355: “Changing Behavior Toward a More
In a series of efforts on developing methods Sustainable Transport System,” in Transport Survey
to impute travel mode from GPS data, the rule-based Methods: Keeping up with a Changing World,
model that was introduced here, performed fairly Emerald Group Publishing Limited, 2009, pp. 621–
well the classification between walk, bike and 634.
motorized means. Motorcycle was the chief culprit [3] J. Wolf et al., Applying GPS Data to Understand
of confusion between bike and motorized groups. Travel Behavior, Volume I: Background, Methods,
The inclusion of motorcycle was a novel and and Tests. Washington, D.C.: Transportation
interesting challenge to mode detection field that has Research Board, 2014.
been around walk, bike, car, bus and train in existing [4] F. Zhao et al., “Exploratory Analysis of a
studies. Although the deterministic method failed to Smartphone-Based Travel Survey in Singapore,”
show very high accuracy and recall levels, it has led Transp. Res. Rec. J. Transp. Res. Board, vol. 2494,
the authors to arrive at two important conclusions. pp. 45–56, Jan. 2015.
First, a more flexible method, a probabilistic for [5] P. Marchal, J.-L. Madre, and S. Yuan,
example, should be used to classify between “Postprocessing Procedures for Person-Based
motorized, walk and car. Second a hierarchical Global Positioning System Data Collected in the
process is promising to reduce the ambiguities French National Travel Survey 2007–2008,”
between modes and thus gain reasonable prediction Transp. Res. Rec. J. Transp. Res. Board, vol. 2246,
results. Specifically, after walk and bike are no. 1, pp. 47–54, Jan. 2011.
separated, more complex techniques (e.g. random [6] P. R. Stopher, V. Daigler, and S. Griffith,
forest) could be used to identify bus, car and “Smartphone app versus GPS Logger: A
motorcycle. comparative study,” Transp. Res. Procedia, vol. 32,
The authors believe that the Hanoi data are pp. 135–145, Jan. 2018.
a valuable resource to implement improvements in [7] S. G. Bricka, S. Sen, R. Paleti, and C. R. Bhat,
the mode detection domain because travel patterns in “An analysis of the factors influencing differences in
Hanoi witness the dominant role of motorcycle and survey-reported and GPS-recorded trips,” Transp.
the slower movement of modes compared with that Res. Part C Emerg. Technol., vol. 21, no. 1, pp. 67–
in industrialized economies, which raises the 88, Apr. 2012.
ambiguity between modes’ behaviors. The following [8] H. Safi, B. Assemi, M. Mesbah, L. Ferreira, and
work is to keep developing models based on fuzzy M. Hickman, “Design and Implementation of a
logic theory and machine learning to do Smartphone-Based Travel Survey,” Transp. Res.
disaggregated classification between five modes, Rec., vol. 2526, no. 1, pp. 99–107, Jan. 2015.
including: walk, bike, bus, motorcycle and car. [9] G. Xiao, Z. Juan, and C. Zhang, “Travel mode
Besides, data are comprised of trip purposes, thus detection based on GPS track data and Bayesian
constructing activity inference algorithms would be networks,” Comput. Environ. Urban Syst., vol. 54,
interesting and feasible research directions. pp. 14–22, Nov. 2015.
[10] Z. Patterson and K. Fitzsimmons, “DataMobile:
Smartphone Travel Survey Experiment,” Transp.

42