Applicability of Willems method for age estimation in Brazilian children: performance of multiple linear regression and artificial neural network

Egyptian Journal of Forensic Sciences

Table 4 Comparison of prediction performance between multivariate linear regression and artificial neural network

Sex	Age interval	ME			MAE			RMSE
Sex	Age interval	OWM	MLR	AAN	OWM	MLR	AAN	OWM	MLR	AAN
F	6–6.99	− 0.42	− 0.59	− 0.48	0.58	0.78	0.51	0.77	0.95	0.63
	7–7.99	− 0.22	− 0.36	− 0.04	0.55	0.61	0.43	0.62	0.71	0.54
	8–8.99	− 0.10	− 0.32	0.01	0.58	0.79	0.66	0.75	0.98	0.77
	9–9.99	− 0.18	− 0.59	− 0.21	0.65	0.90	0.72	0.83	1.13	0.96
	10–10.99	− 0.06	− 0.60	− 0.19	0.67	0.91	0.84	0.86	1.13	1.04
	11–11.99	− 0.09	− 0.73	− 0.48	0.62	0.91	0.94	0.83	1.09	1.08
	12–12.99	0.26	− 0.30	− 0.08	0.98	0.78	0.79	1.23	0.99	1.07
	13–13.99	− 0.25	− 0.24	− 0.27	1.01	0.66	0.82	1.29	0.79	0.91
	14–14.99	0.46	0.54	0.54	1.23	0.70	0.68	1.43	0.89	1.03
	15–15.99	0.71	1.17	1.20	1.05	1.17	1.20	1.32	1.28	1.32
M	6–6.99	− 0.54	− 0.06	− 0.39	0.63	0.52	0.45	0.74	0.63	0.52
	7–7.99	− 0.39	0.05	0.01	0.52	0.49	0.45	0.65	0.66	0.54
	8–8.99	− 0.29	0.09	− 0.15	0.64	0.66	0.69	0.81	0.83	0.87
	9–9.99	− 0.36	− 0.08	− 0.27	0.79	0.89	0.80	0.95	1.04	0.96
	10–10.99	− 0.14	0.06	− 0.15	0.73	0.81	0.72	0.89	1.07	0.93
	11–11.99	− 0.33	− 0.17	− 0.29	0.74	0.73	0.73	0.93	0.89	0.88
	12–12.99	0.00	0.17	− 0.03	0.80	0.88	0.78	0.99	1.06	0.99
	13–13.99	− 0.13	0.17	0.01	0.93	0.74	0.75	1.14	1.01	0.94
	14–14.99	− 0.17	0.42	0.16	1.04	0.60	0.63	1.25	0.85	0.86
	15–15.99	0.54	1.35	1.00	0.98	1.35	1.00	1.33	1.53	1.27

F females, M males, OWM original Willems model, MLR multivariate linear regression, ANN artificial neural network, ME mean error, MAE mean absolute error, RMSE root mean squared error
Age interval expressed in years