An Explainable Stacked Ensemble Model for Static Route-Free Estimation of Time of Arrival

Table 3

Comparison of our ETA models of the first and second level based on different evaluation metrics.

	Data set	New York city			Washington DC
	Evaluation metric	MAE (seconds)	MRE	MAPE	MAE (seconds)	MRE	MAPE
Level 1	L1-RF	180.694	0.2158	27.8689	179.5912	0.2373	30.1512
	L1-XGBoost	183.4192	0.219	27.1137	190.2613	0.2514	30.4033
	L1-FCNN	178.2321	0.2129	23.7561	169.8152	0.2244	24.372
Level 2	L2-MLR	172.2439	0.2057	25.2758	171.178	0.2262	27.1985
	L2-RF	183.2319	0.2188	26.9828	183.7377	0.2428	29.5762
	L2-XGBoost	173.6526	0.2074	25.3077	172.7287	0.2283	27.5419
	L2-FCNN	169.4285	0.2023	22.9121	167.9959	0.222	24.6133
Baselines	[12]	185.9265	0.2256	23.8429	181.1275	0.2374	27.4261
	[1]	201.5998	0.2455	28.1508	203.8581	0.2673	35.4898
	[13]	185.3999	0.2257	28.3598	174.3907	0.2286	25.7570

This prediction precision is better than the one presented by Schleibaum et al. [4].