Table 5 Summary of the performance of the algorithms for the OA group, considering input 1 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)	0.28	9.43	0.36	0.47	0.22	\(-\) 0.22	\(-\) 0.36	\(-\) 0.08	0.59	15.35	0.78	0.17	0.03	\(-\) 0.58	\(-\) 0.83	\(-\) 0.33
(2) Ensemble trees (LSBoost)	0.46	15.61	0.52	0.27	0.08	\(-\) 0.04	\(-\) 0.29	0.20	0.50	14.20	0.73	0.58	0.34	\(-\) 0.46	\(-\) 0.73	\(-\) 0.18
(3) Linear SVR	0.34	11.19	0.40	0.85	0.72	0.34	0.24	0.44	0.69	19.81	0.81	0.65	0.42	\(-\) 0.67	\(-\) 0.89	\(-\) 0.46
(4) Quadratic SVR	0.57	19.08	0.66	0.62	0.39	\(-\) 0.50	\(-\) 0.71	\(-\) 0.29	0.66	18.16	0.79	0.57	0.33	\(-\) 0.66	\(-\) 0.87	\(-\) 0.45
(5) Cubic SVR	0.42	13.93	0.48	0.59	0.35	\(-\) 0.37	\(-\) 0.52	\(-\) 0.22	1.27	41.48	1.94	0.74	0.55	\(-\) 1.23	\(-\) 1.95	\(-\) 0.50
(6) Gaussian SVR	0.18	6.13	0.21	0.85	0.72	\(-\) 0.16	\(-\) 0.22	\(-\) 0.09	0.52	14.79	0.58	0.14	0.02	\(-\) 0.26	\(-\) 0.51	\(-\) 0.01
(7) Linear regression	0.17	5.79	0.21	0.84	0.70	0.13	0.05	0.21	0.75	20.26	0.91	0.18	0.03	\(-\) 0.70	\(-\) 0.98	\(-\) 0.42
(8) Lasso regression	0.17	5.63	0.24	0.68	0.47	0.11	0.01	0.21	0.71	19.57	0.84	0.45	0.20	\(-\) 0.68	\(-\) 0.92	\(-\) 0.44
(9) Ridge regression	0.28	9.61	0.41	0.54	0.29	0.25	0.10	0.41	0.63	17.56	0.75	0.62	0.39	\(-\) 0.61	\(-\) 0.82	\(-\) 0.40
(10) Binary decision tree	0.38	13.00	0.45	0.71	0.51	\(-\) 0.37	\(-\) 0.49	\(-\) 0.24	0.52	13.82	0.65	0.09	0.01	\(-\) 0.33	\(-\) 0.60	\(-\) 0.06
(11) GR (K.-exponential)	0.15	5.11	0.18	0.84	0.70	\(-\) 0.10	\(-\) 0.17	\(-\) 0.02	0.51	13.57	0.63	0.28	0.08	\(-\) 0.38	\(-\) 0.62	\(-\) 0.14
(12) GR (K.-squared exponential)	0.26	8.57	0.28	0.85	0.72	\(-\) 0.24	\(-\) 0.31	\(-\) 0.18	0.52	14.52	0.59	0.08	0.01	\(-\) 0.26	\(-\) 0.51	0.00
(13) GR (K.-matern 32)	0.12	4.13	0.14	0.92	0.86	\(-\) 0.10	\(-\) 0.15	\(-\) 0.05	0.53	14.80	0.61	0.02	0.00	\(-\) 0.30	\(-\) 0.56	\(-\) 0.04
(14) GR (K.-matern 52)	0.22	7.40	0.24	0.87	0.76	\(-\) 0.21	\(-\) 0.27	\(-\) 0.15	0.53	14.89	0.61	0.02	0.00	\(-\) 0.28	\(-\) 0.54	\(-\) 0.02
(15) GR (K.-rational quadratic)	0.21	7.18	0.23	0.87	0.76	\(-\) 0.20	\(-\) 0.26	\(-\) 0.14	0.53	14.79	0.60	0.01	0.00	\(-\) 0.27	\(-\) 0.53	\(-\) 0.02
(16) ETSVR-Kernel Linear	0.40	13.63	0.56	0.53	0.28	0.39	0.20	0.59	0.67	18.59	0.80	0.59	0.35	\(-\) 0.67	\(-\) 0.88	\(-\) 0.45
(17) Kernel ridge regression	0.44	14.88	0.59	0.55	0.30	0.43	0.24	0.63	0.66	18.46	0.79	0.56	0.32	\(-\) 0.64	\(-\) 0.86	\(-\) 0.41
(18) Nyström ridge regression	0.53	17.66	0.71	0.49	0.24	0.51	0.28	0.75	0.35	9.29	0.55	0.51	0.26	\(-\) 0.26	\(-\) 0.49	\(-\) 0.03
(19) DNNE	0.35	11.61	0.43	0.74	0.55	0.28	0.12	0.43	0.17	4.73	0.22	0.95	0.90	\(-\) 0.15	\(-\) 0.23	\(-\) 0.07
(20) kNN weighted mean	0.32	10.38	0.37	0.65	0.43	\(-\) 0.31	\(-\) 0.40	\(-\) 0.23	0.50	13.21	0.66	0.14	0.02	\(-\) 0.39	\(-\) 0.65	\(-\) 0.14
(21) RKNNWTSVR	0.36	12.14	0.49	0.61	0.38	0.35	0.19	0.51	0.71	19.82	0.83	0.54	0.30	\(-\) 0.69	\(-\) 0.92	\(-\) 0.46
(22) LTSVR	0.44	14.79	0.64	0.43	0.18	0.42	0.19	0.65	0.56	15.78	0.69	0.66	0.43	\(-\) 0.55	\(-\) 0.75	\(-\) 0.35
(23) Stepwise glm	0.17	5.60	0.23	0.71	0.50	\(-\) 0.16	\(-\) 0.24	\(-\) 0.08	0.61	17.14	0.73	0.67	0.45	\(-\) 0.61	\(-\) 0.80	\(-\) 0.41
(24) Neural networks	0.24	8.00	0.28	0.77	0.59	0.06	\(-\) 0.07	0.19	0.79	20.56	1.02	0.26	0.07	\(-\) 0.77	\(-\) 1.10	\(-\) 0.45

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 5 Summary of the performance of the algorithms for the OA group, considering input 1 as training data.

Search

Quick links