Page 134 - 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

human knowledge and reasoning into travel economic characteristics of interviewees were
decisions is the focus of this study. collected regarding location, gender, age, education,
The main objective of this study is to investigate occupation, personal income, and household
the possibility of two AI methods namely EDT income. They also asked for their travel behavior
Bagged and SVM to predict the travel decisions of characterized by mobility-related factors including
transport users. For the development of models, trip mode, travel distance, trip purpose, and the
travel interview survey was conducted with the possible impact of parking charge on their travel
involvement of 311 transport users at different land- decisions. All transport users were given five travel
use types in Hanoi, Vietnam. A number of 933 data alternatives, including (1) using current mode, (2)
samples were then collected from 311 transport users shift to bus, (3) shift to walk, (4) select another
and discretized to construct the two AI “black- destination, and (5) shift to taxi.
boxes”. Validation of models was conducted using Travel decisions also strongly influenced by
various criteria namely confusion matrix, Root Mean parking characteristics, for instance, walking time
Square Error (RMSE), Mean Absolute Error (MAE) and walking distance from parking location to the
and accuracy. To finely estimate the robustness of destination [39], provision of free parking [41],
the two proposed AI algorithms, 1000 Monte Carlo parking requirements [42], strategies for parking
simulations were then performed. [43], parking charge [44], and parking fee scheme
[45]. Among those influences, parking charge has
2. Data and Methods been used as an effective instrument for traffic
management [46]–[48]. In order to emphasize the
This section describes the data used for this
study, the classifiers used for modeling travel decision in selecting travel alternatives under the
decisions, and the methodology for evaluating their impact of parking charge, each transport users was
performance. asked three times with the increased level of
proposed parking fee: (1) pay at the current rate
2.1. Impact Factors of Travel Decisions and Travel (100%), two times higher (200%), and three times
Interview Survey higher (300%). Many people in Hanoi do not pay for
their parking vehicles. For instance, shoppers have
Travel decisions are the accumulation of free parking whenever they buy something in the
individuals’ preferences influenced by many impact shops, or the parking fees of commuters are paid by
factors which relate to socio-economic their companies.
characteristics of travelers, including gender [32] After being carefully collected, a number of 933
occupation [33], [34], income [33], [35], trip cost data samples was obtained from 311 transport users
[34], [36], travel distance [37], travel mode and with three times asking for willingness-to-pay for the
departure time [38], the quality of walkway [39], and parking charge. The preprocessing data procedure
car ownership rate [37], [40]. was then carried out to construct the input and output
To analyze the impact of such factors on travel spaces for the two AI numerical tools.
decisions, travel interview survey of transport users
was conducted in June 2016 in three major zones in 2.2. Preparation of Datasets
Hanoi, including Old Quarter Area, Developed Area,
and New Development Area. Old Quarter Area is In the present study, we consider input
characterized by ancient buildings. It is the location parameters including location (X1), gender (X2),
of the busiest shopping streets and tourist attractions. age (X3), education (X4), occupation (X5), personal
Developed Area is located inside the Ring Road II income (X6), household income (X7), trip mode
and having the highest population density. There is (X8), trip purpose (X9), trip length (X10), trip cost
mixed land use area, with the combination of (X11), on-vehicle duration (X12), parking and
Government authorities, office buildings and walking duration (X13), real parking charge (X14)
shopping streets. New Development Area is placed and proposed parking charge (X15), respectively.
inside the area of Ring Road II and Ring Road III The output (Y), referred to the travel decisions of
with modern condominiums, office buildings and transport users, was discretized in five alternatives as
universities. mentioned above. To generate the datasets for
A total of 311 people was randomly selected at modeling, 933 data samples collected was divided
the buildings of different land use types. The socio-

109