Self-Tuning Inference Model for Settlement in Shield Tunneling: A Case Study of the Taipei Mass Rapid Transit System’s Songshan Line

Table 9

Comparison with data balanced by SMOTE.

Algorithm	Training						Testing
	MAPE (%)	RMSE	MAE	R	RI	Rank	MAPE (%)	RMSE	MAE	R	RI	Rank
Original dataset
BPNN	10.75	2.2276	1.4593	0.9023	0.7201	4	12.16	2.3635	1.6551	0.8874	0.7844	3
LSSVM	8.49	1.9266	1.3073	0.9648	0.8250	2	10.12	2.2621	1.5159	0.9442	0.8322	2
ELSIM	8.49	1.9088	1.3308	0.9264	0.7885	3	11.59	5.5700	1.8920	0.7807	0.6331	4
SVM	25.84	6.7033	4.8161	0.8136	0.1090	5	32.51	7.1774	5.6747	0.1400	0.0000	5
SOS-LSSVM	5.00	1.2579	0.8031	0.9811	0.9188	1	7.63	2.0467	1.2002	0.9567	0.8828	1
SMOTE oversampling
BPNN	5.53	1.5974	0.9583	0.9845	0.9175	4	5.84	1.7868	1.0125	0.9809	0.9691	2
LSSVM	5.32	1.4877	0.8906	0.9873	0.9317	3	6.29	1.8163	1.0116	0.9806	0.9662	3
ELSIM	3.49	1.1910	0.6035	0.9850	0.9529	2	5.24	2.9411	1.0921	0.8949	0.9078	4
SVM	29.71	5.6932	4.1081	0.8697	0.0000	5	32.81	6.4231	4.8021	0.8062	0.0000	5
SOS-LSSVM	2.61	0.8461	0.4467	0.9950	1.0000	1	4.15	1.4436	0.6654	0.9879	1.0000	1
SMOTE median sampling
BPNN	4.11	1.4605	0.8513	0.9823	0.8995	3	4.39	1.5473	0.9093	0.9802	0.9540	2
LSSVM	4.50	1.4727	0.9336	0.9830	0.8907	4	5.09	1.7091	1.0575	0.9759	0.9293	3
ELSIM	3.44	1.2499	0.7272	0.9956	0.9469	2	4.69	1.9611	0.9669	0.9322	0.8862	4
SVM	25.60	6.2592	4.3272	0.8483	0.0000	5	34.69	6.7349	5.3017	0.7116	0.0000	5
SOS-LSSVM	2.09	0.8341	0.4202	0.9943	0.9977	1	3.09	1.2293	0.6277	0.9871	1.0000	1