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

Table 5

Sensitivity and specificity.

(a) Disease-free survival.

	LR	NN-varLR	NN-varNN
Sensitivity	71.5	82.2	80.3
Specificity	68.9	77.9	79.3
False negative rate	28.5	17.8	19.7
False positive rate	31.1	22.1	20.7
Positive predictive value	11.9	18.0	18.6

(b) Mortality from cancer causes.

	LR	NN-varLR	NN-varNN
Sensitivity	80.5	86.7	87.5
Specificity	77.9	76.0	77.3
False negative rate	19.5	13.3	12.5
False positive rate	22.1	24.0	22.7
Positive predictive value	17.4	17.2	18.2

(c) Local recurrence.

	LR	NN-varLR	NN-varNN
Sensitivity	68.6	72.4	72.9
Specificity	65.1	66.7	64.9
False negative rate	31.4	27.6	27.1
False positive rate	34.9	33.3	35.1
Positive predictive value	8.3	9.1	8.8

(d) Metastatic recurrence.

	LR	NN-varLR	NN-varNN
Sensitivity	71.3	70.6	71.9
Specificity	78.1	78.8	78.5
False negative rate	28.7	29.4	28.1
False positive rate	21.9	21.2	21.5
Positive predictive value	25.9	26.4	26.4