Page 64 - 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

Several installed the app but forgot to turn off save research goals only and they should not published or
mode and/or turn on location tracking, thus their shared in any cases. Several users claimed technical
phones did not collect data at all. Among of 72, 9 challenges to split and merge segments.
have not made any validation yet. So, data of 63
persons completing the surveys were included in the 4. Mode detection method
further analyses. Their data included 795 days. On 4.1. Data preparation
average, a person has 12.6 days that is much higher Before implementing mode detection, bad
the minimum number of 7 days indicated above, GPS points were eliminated according to two rules.
which emphasized participants’ responsibility and First, a point will be invalid if it has the same
enthusiasm for the survey. timestamp or coordinate as the previous. Second, a
Data were delivered to DEST in JSON files point will be deleted if its instantaneous speed is over
and structured into segments. A segment assigned an 120 km/h that is the maximum speed allowed on the
ID number includes ID of phone collecting it and a road in Vietnam. The distance between points was
series of GPS points. A point encompasses latitude, calculated by equations of Vincenty [22].
longitude and timestamp. Points were collected at In the data of 63 persons, there were a
high frequency ranging between 1 and 3 seconds. number of segments abroad (in Philippines, China
3.2 Respondents’ views on TRavelVU and Japan) and outside Hanoi. The authors removed
At the end of surveys, DEST listened to all of them except from domestic ones connecting
feedbacks of participants about both advantages and between Hanoi and other provinces. Finally, 2791
disadvantages of the surveys in general and the app segments on foot, by bike, bus, motor and car were
in particular. Most of all showed their happiness and valid for mode detection. There was not any segment
interest in the mobility survey using smartphone. of metro because its kick-off delayed at the end of
Validation process gave them a review about their 2019. The mode share presented in Figure 3 is
daily travel patterns, especially how to make active compatible with travel patterns Hanoi where
travel. They appreciated the app’s friendly display motorcycle is major whilst bus and bike are auxiliary
with pictures for choices that allowed them to [23], [24].
validate data with a limited English level. They also
agreed with the great potential for using smartphone 587, 21% 758, 27%
than self-reported investigations to collect mobility
data.
Nevertheless, some especially who traveled
a lot, complained about the considerable battery 104, 4%
consumption of the app. High drainage came from 97, 3%
the wish of collecting high-resolution data at very
short intervals (1-3 s). Most of participants felt
unhappy with the wrong recommendation of modes, 1245, 45%
which made them to correct modes of many
segments. This problem had anticipated before. Walk Bike Bus Motor Car
Because the app utilizes the thresholds calibrated
from travel patterns in Sweden and other countries Fig. 3 Mode share in the sample
in the Global North, many motorcycle segments 4.2 Mode detection method
were proposed wrongly as bike that gain the Here, the target is to distinguish between
maximum speed up to 40 Km/h [11], [21] on cycle walk, bike and motorized modes by a deterministic
lanes. Language barrier was another source of method.
unhappiness. Some declared that during the To determine rules, the authors sought the
installation process, the app launched instructions thresholds to detect walk, bike and motorized modes
and questions in Swedish and they had to change to in the literature. The study in the Netherlands [11]
English after the app was usable in their phones. applied the maximum speed and the average speed
However, this was not a big problem because they of 14 km/h and 10 km/h, respectively for
requested and received prompt supports from the identification of walk. In similar way albeit with
corresponding staff of DEST. Privacy concern was high levels of 45 km/h and 25 km/h, bike ones were
paid close and great attention. The vast majority of detected. The rest, of course was motorized
participants emphasized that their data were used for counterparts. Stopher et al. (2008) [21] used lower

39