Design of a Clinical Decision Support System for Fracture Prediction Using Imbalanced Dataset

Table 1

Predictive performance by randomly undersampling the imbalanced training dataset.

Training subset	AC (%)	SE (%)	SP (%)	G-mean	AUC
1	66.32	63.02	66.53	0.6484	0.704
2	65.67	63.86	65.78	0.6481	0.698
3	67.75	60.92	68.18	0.6445	0.704
4	65.07	64.28	65.12	0.6470	0.700
5	65.42	62.18	65.62	0.6388	0.705
6	65.78	62.18	66.10	0.6411	0.713
7	67.75	60.92	68.18	0.6445	0.710
8	65.57	62.18	65.78	0.6396	0.710
9	67.29	65.54	67.29	0.6641	0.716
10	65.12	63.44	65.23	0.6433	0.709
Average	66.17	62.85	66.38	0.6439	0.707