Table 3 Predictive accuracy of the k-NN model for different choices of the number of nearest neighbors × 500 replications and comparison with conventional logistic regression. Accuracy selects the best number of neighbors within a larger range and uses predictor importance when calculating distances
From: Predicting complete cytoreduction for advanced ovarian cancer patients using nearest-neighbor models
Number of nearest neighbors | Mean predictive accuracy (%) | Minimum predictive accuracy (%) | Maximum predictive accuracy (%) | Mean accuracy of TPs (%) | Mean accuracy of TNs (%) | Mean accuracy of FPs (%) | Mean accuracy of FNs (%) |
3 | 57.5 | 41.2 | 72.5 | 37.0 | 70.7 | 29.3 | 63.0 |
4 | 57.8 | 39.2 | 78.4 | 38.2 | 70.6 | 29.4 | 61.8 |
5 | 60.7 | 43.1 | 74.5 | 36.4 | 76.3 | 23.7 | 63.6 |
6 | 60.8 | 41.2 | 82.4 | 37.5 | 75.8 | 24.2 | 62.5 |
7 | 62.9 | 45.1 | 78.4 | 35.2 | 80.6 | 19.4 | 64.8 |
8 | 62.8 | 43.1 | 78.4 | 35.7 | 80.2 | 19.8 | 64.3 |
9 | 64.5 | 47.1 | 80.4 | 34.5 | 83.6 | 16.4 | 65.5 |
10 | 64.4 | 47.1 | 78.4 | 35.0 | 83.2 | 16.8 | 65.0 |
11 | 65.2 | 45.1 | 78.4 | 34.6 | 84.7 | 15.3 | 65.4 |
12 | 64.7 | 43.1 | 82.4 | 34.8 | 83.9 | 16.1 | 65.2 |
13 | 65.5 | 45.1 | 80.4 | 34.1 | 85.6 | 14.4 | 65.9 |
14 | 65.3 | 45.1 | 80.4 | 33.7 | 85.5 | 14.5 | 66.3 |
15 | 65.8 | 47.1 | 82.4 | 32.5 | 87.1 | 12.9 | 67.5 |
16 | 65.4 | 47.1 | 80.4 | 32.0 | 86.8 | 13.2 | 68.0 |
17 | 65.5 | 49.0 | 80.4 | 30.5 | 88.0 | 12.0 | 69.5 |
18 | 65.5 | 47.1 | 76.5 | 29.8 | 88.3 | 11.7 | 70.2 |
19 | 65.8 | 47.1 | 80.4 | 28.8 | 89.4 | 10.6 | 71.2 |
20 | 65.6 | 43.1 | 80.4 | 28.1 | 89.5 | 10.5 | 71.9 |
Logistic regression results | Mean predictive accuracy (%) | Minimum predictive accuracy (%) | Maximum predictive accuracy (%) | Mean accuracy of TPs (%) | Mean accuracy of TNs (%) | Mean accuracy of FPs (%) | Mean accuracy of FNs (%) |
– | 63.4 | 2.0 | 80.4 | 42.7 | 76.7 | 23.1 | 57.1 |