Page 143 - 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

Bagged exhibit a large order of perturbation when optimal numbers of Monte Carlo simulations in
the number of Monte Carlo simulations is small (N order to obtain the stationary result in case of SVM
< 10). The fluctuation in several cases is important, are Nopt = 300, 400 and 20 in terms of RMSE, MAE
e.g. 9% for MAE of testing EDT Bagged, or 5% in and Accuracy, respectively. On the other hand, the
the cases of RMSE for both EDT Bagged and SVM. optimal number of Monte Carlo simulations in case
It is noticed that the discontinuous lines in Figure 7 of EDT Bagged are Nopt = 120, 220 and 20 in terms
represent an interval of ±1% of variation around the of RMSE, MAE and accuracy, respectively. It can be
stationary result, the optimal number of Monte Carlo deduced that EDT Bagged reached the stationary
simulation is defined as the number of runs when the solutions with a smaller number of Nopt, showing
curves were totally inside the ±1% range. The the robustness of the model compared to SVM.

Figure 7. Statistical graphs of different Discontinuous blue lines represent interval
criteria for 1000 Monte Carlo simulations of ±1% of variation around the stationary
in case of: (a) RMSE for testing SVM; (b) result.
RMSE for testing EDT Bagged; (c) MAE
for testing SVM; (d) MAE for testing EDT The probability density distributions of SVM
Bagged; (e) Accuracy for testing SVM; (f) and EDT Bagged algorithm in terms of RMSE, MAE
Accuracy for testing EDT Bagged. and accuracy are represented in Figure 8. It can be

118