Page 142 - Proceeding of Atrans Young Researcher's Forum 2019_Neat
P. 142
“Transportation for A Better Life:
Smart Mobility for Now and Then”
23 August 2019, Bangkok, Thailand
Figure 6. Error distribution for 1000 Statistical results of the two AI models are
Monte Carlo simulation in case: (a) RMSE presented in Figure 7 for RMSE, MAE and Accuracy
for testing SVM; (b) RMSE for testing of testing SVM and EDT Bagged. In this study, the
EDT Bagged; (c) MAE for testing SVM; convergence estimator was defined as the following
(d) MAE for testing EDT Bagged; (e) equation [65], [66]:
Accuracy for testing SVM; (f) Accuracy
for testing EDT Bagged;
1 N
N
N a f = W , (10)
convergence
NW k=1 k
where N is defined as the number of Monte solution of the random variable of interest. Once
Carlo simulations, W is the mean value of the determined, for instance the optimal number of
random variable W (in this case, they are RMSE, Monte Carlo simulation Nopt, reliable statistical
MAE and Accuracy). The statistical convergence analysis of the considered random variable could be
function is helpful to investigate the stationary obtained. In terms of all criteria, both SVM and EDT
117
