Comparison of Artificial Neural Network with Logistic Regression as Classification Models for Variable Selection for Prediction of Breast Cancer Patient Outcomes

Table 2

Details of variables selections for “specific mortality”.

	Forward		Forward		Backward		Backward		Genetic		Genetic		Global
	0.0001		0.0		0.0001		0.0		0.0001		0.0
	#	%	#	%	#	%	#	%	#	%	#	%	#	%
Side of Breast	0	0%	4	20%	0	0%	4	20%	0	0%	4	20%	12	10%
History of Cancer	0	0%	11	55%	0	0%	10	50%	1	5%	4	20%	26	22%
Menopausal	16	80%	19	95%	19	95%	17	85%	17	85%	19	95%	107	89%
Cutaneous inflam	0	0%	4	20%	0	0%	3	15%	0	0%	3	15%	10	8%
Skin invasion	6	30%	14	70%	3	15%	11	55%	5	25%	13	65%	52	43%
Clin. Nb. Nodules	20	100%	20	100%	20	100%	20	100%	20	100%	20	100%	120	100%
Invaded Nodes	20	100%	19	95%	20	100%	19	95%	20	100%	20	100%	118	98%
Stade N	16	80%	19	95%	18	90%	18	90%	14	70%	20	100%	105	88%
Nb of tum (histo)	8	40%	16	80%	6	30%	20	100%	5	25%	18	90%	73	61%
Size of tum (histo)	0	0%	4	20%	0	0%	5	25%	0	0%	3	15%	12	10%
SBR Grade	19	95%	20	100%	19	95%	20	100%	20	100%	20	100%	118	98%
Lymphatic Embolus	14	70%	15	75%	13	65%	18	90%	13	65%	14	70%	87	73%
Nervous spread	0	0%	11	55%	4	20%	7	35%	1	5%	6	30%	29	24%
Limits Exeresis	0	0%	10	50%	0	0%	8	40%	0	0%	5	25%	23	19%
Skin infiltating	2	10%	9	45%	7	35%	8	40%	0	0%	11	55%	37	31%
Skin embolus	0	0%	6	30%	0	0%	2	10%	0	0%	2	10%	10	8%
Nipple infiltrating	1	5%	4	20%	0	0%	4	20%	0	0%	3	15%	12	10%
Nipple Cancer in situ	0	0%	4	20%	0	0%	1	5%	0	0%	1	5%	6	5%
Nipple embolus	2	10%	5	25%	0	0%	3	15%	1	5%	3	15%	14	12%
Nipple Paget	0	0%	9	45%	0	0%	10	50%	0	0%	5	25%	24	20%
Pectoral muscle inv	0	0%	5	25%	0	0%	3	15%	0	0%	3	15%	11	9%
Age	6	30%	18	90%	5	25%	19	95%	7	35%	19	95%	74	62%
Time diag-firstTreat	0	0%	2	10%	0	0%	2	10%	0	0%	0	0%	4	3%
Time diag-firstSurgery	0	0%	8	40%	0	0%	11	55%	0	0%	6	30%	25	21%
Histology	19	95%	19	95%	20	100%	20	100%	20	100%	20	100%	118	98%
Necrosis	11	55%	16	80%	13	65%	16	80%	15	75%	13	65%	84	70%
Infiltrating tumour	0	0%	8	40%	0	0%	11	55%	1	5%	11	55%	31	26%
Clinical Size	20	100%	20	100%	19	95%	20	100%	20	100%	19	95%	118	98%
Capsular breaking	18	90%	19	95%	12	60%	19	95%	17	85%	18	90%	103	86%
Oestro receptors	19	95%	18	90%	17	85%	20	100%	17	85%	16	80%	107	89%
Progest receptors	20	100%	20	100%	20	100%	20	100%	20	100%	20	100%	120	100%
Nb. Nodes Invad	20	100%	20	100%	20	100%	20	100%	20	100%	20	100%	120	100%