Table 7 Summary of the performance of the algorithms for the OA group, considering Input 3 as training data.

From: On the prediction of tibiofemoral contact forces for healthy individuals and osteoarthritis patients during gait: a comparative study of regression methods

Function	1st Knee contact peak (N/body weight)								2nd Knee contact peak (N/body weight)
Function	MAE	RPE	RMSE	R	R2	MDF	LCI	UCI	MAE	RPE	RMSE	R	R2	MDF	LCI	UCI
(1) Ensemble trees (bagging)	0.35	11.99	0.42	0.69	0.48	0.30	0.16	0.44	0.55	14.07	0.73	0.13	0.02	\(-\) 0.51	\(-\)0.76	\(-\)0.26
(2) Ensemble trees (LSBoost)	0.25	8.55	0.37	0.44	0.20	\(-\)0.02	\(-\)0.19	0.16	0.73	19.95	0.91	0.02	0.00	\(-\)0.65	\(-\)0.96	\(-\)0.34
(3) Linear SVR	0.07	2.48	0.11	0.91	0.82	\(-\)0.01	\(-\)0.06	0.04	0.21	5.80	0.26	0.96	0.92	\(-\) 0.20	\(-\) 0.28	\(-\) 0.13
(4) Quadratic SVR	0.48	15.64	0.59	0.81	0.66	0.16	\(-\)0.12	0.43	0.27	7.42	0.34	0.93	0.87	\(-\)0.18	\(-\)0.32	\(-\)0.05
(5) Cubic SVR	0.15	5.10	0.21	0.86	0.74	0.14	0.07	0.21	0.25	6.66	0.32	0.94	0.88	\(-\)0.18	\(-\)0.30	\(-\)0.05
(6) Gaussian SVR	0.09	3.16	0.11	0.92	0.85	\(-\)0.03	\(-\)0.09	0.02	0.27	7.26	0.33	0.93	0.86	\(-\)0.18	\(-\)0.31	\(-\)0.04
(7) Linear regression	1.11	36.04	1.32	0.84	0.70	1.10	0.75	1.45	0.53	13.86	0.67	0.44	0.19	\(-\)0.47	\(-\)0.70	\(-\)0.25
(8) Lasso regression	0.10	3.51	0.16	0.84	0.71	0.07	0.00	0.14	0.68	18.10	0.84	0.34	0.12	\(-\)0.67	\(-\)0.92	\(-\)0.43
(9) Ridge regression	0.17	5.80	0.23	0.89	0.79	0.17	0.10	0.24	0.47	12.80	0.57	0.77	0.60	\(-\)0.47	\(-\)0.63	\(-\)0.31
(10) Binary decision tree	0.17	5.77	0.22	0.72	0.51	0.02	\(-\)0.08	0.13	0.75	19.96	0.92	0.09	0.01	\(-\)0.69	\(-\)0.98	\(-\)0.39
(11) GR (K.-exponential)	0.12	4.16	0.15	0.89	0.79	0.06	\(-\)0.01	0.12	0.54	14.12	0.71	0.53	0.28	\(-\)0.54	\(-\)0.76	\(-\)0.32
(12) GR (K.-squared exponential)	0.10	3.43	0.12	0.93	0.86	0.00	\(-\)0.05	0.06	0.34	9.17	0.42	0.92	0.84	\(-\)0.25	\(-\)0.42	\(-\)0.09
(13) GR (K.-matern 32)	0.10	3.34	0.11	0.93	0.87	\(-\) 0.02	\(-\) 0.07	0.04	0.34	8.99	0.44	0.92	0.85	\(-\)0.31	\(-\)0.46	\(-\)0.15
(14) GR (K.-matern 52)	0.11	3.88	0.15	0.92	0.84	\(-\)0.07	\(-\)0.13	\(-\)0.01	0.34	9.12	0.43	0.92	0.85	\(-\)0.28	\(-\)0.44	\(-\)0.13
(15) GR (K.-rational quadratic)	0.10	3.43	0.12	0.93	0.86	\(-\)0.04	\(-\)0.09	0.02	0.34	9.17	0.42	0.92	0.84	\(-\)0.25	\(-\)0.42	\(-\)0.09
(16) ETSVR-Kernel linear	0.11	3.53	0.16	0.85	0.72	\(-\)0.06	\(-\)0.13	0.01	0.36	9.84	0.43	0.89	0.79	\(-\)0.35	\(-\)0.47	\(-\)0.23
(17) Kernel ridge regression	0.10	3.19	0.14	0.88	0.77	\(-\)0.05	\(-\)0.11	0.01	0.41	11.15	0.50	0.85	0.71	\(-\)0.40	\(-\)0.54	\(-\)0.27
(18) Nyström ridge regression	0.24	7.80	0.31	0.53	0.28	0.00	\(-\)0.15	0.14	0.61	17.07	0.77	0.60	0.36	\(-\)0.59	\(-\)0.83	\(-\)0.36
(19) DNNE	0.69	22.91	0.82	0.67	0.45	0.68	0.46	0.90	0.57	15.35	0.69	0.04	0.00	\(-\)0.32	\(-\)0.61	\(-\)0.03
(20) kNN weighted mean	0.31	10.50	0.41	0.73	0.53	0.30	0.16	0.43	0.59	15.33	0.78	0.03	0.00	\(-\)0.54	\(-\)0.81	\(-\)0.27
(21) RKNNWTSVR	0.11	3.50	0.15	0.87	0.76	\(-\)0.01	\(-\)0.09	0.06	0.39	10.73	0.47	0.86	0.75	\(-\)0.38	\(-\)0.51	\(-\)0.25
(22) LTSVR	0.28	8.89	0.34	0.49	0.24	\(-\)0.27	\(-\)0.37	\(-\)0.17	0.62	16.98	0.75	0.61	0.37	\(-\)0.62	\(-\)0.82	\(-\)0.42
(23) Stepwise glm	0.12	3.84	0.15	0.84	0.70	\(-\)0.08	\(-\)0.14	\(-\)0.01	0.58	15.47	0.71	0.65	0.42	\(-\)0.58	\(-\)0.77	\(-\)0.38
(24) Neural networks	0.15	5.06	0.19	0.88	0.77	\(-\)0.07	\(-\)0.15	0.01	0.15	4.23	0.17	0.95	0.91	\(-\)0.06	\(-\)0.14	0.01

Best results (i.e. highest accuracy) are in bold.
RMSE root mean squared error, R Pearson correlation coefficient, (R\(^2\)) the coefficient of determination, MDF mean delta force, LCI lower confidence interval, UPF upper confidence interval, GR Gaussian regression, K Kernel.

Back to article page

Table 7 Summary of the performance of the algorithms for the OA group, considering Input 3 as training data.

Search

Quick links