Table 3.

Impact of gender on the performance of eye colour prediction models developed with a Polish sample set using neural networks and multinomial logistic regression methods

Eye colour category	Prediction accuracy	Mathematical method
		Neural networks		Multinomial logistic regression
		Gender not included	Gender included	Gender not included	Gender included
Blue eye colour	AUC	0.889	0.863	0.872	0.880
	Sensitivity [%]	94.06	92.51	93.11	93.11
	Specificity [%]	74.34	73.85	74.07	74.07
Green eye colour	AUC	0.667	0.709	0.611	0.628
	Sensitivity [%]	0.71	0.00	0.00	0.00
	Specificity [%]	99.55	99.88	99.88	99.88
Hazel eye colour	AUC	0.833	0.843	0.797	0.800
	Sensitivity [%]	64.88	63.60	65.20	65.20
	Specificity [%]	81.04	79.91	80.22	80.22
Brown eye colour	AUC	0.917	0.918	0.889	0.892
	Sensitivity [%]	34.23	35.04	33.80	33.80
	Specificity [%]	95.54	94.18	94.52	94.52