Table 3.
Fisher discriminant analysis and cross-validation
Subtypes/n | Training dataset | Prediction accuracy | |||||
---|---|---|---|---|---|---|---|
1 | 2 | 3 | 4 | 5 | |||
Fisher discriminant analysis | 1 | 28 | 0 | 1 | 0 | 0 | 97% |
2 | 0 | 31 | 0 | 0 | 0 | 100% | |
3 | 0 | 0 | 16 | 0 | 1 | 94% | |
4 | 1 | 1 | 0 | 22 | 0 | 92% | |
5 | 0 | 0 | 0 | 0 | 11 | 100% | |
Total | – | – | – | – | – | – | 96.4% |
Leave-one- out cross- validation | 1 | 23 | 4 | 0 | 0 | 2 | 79% |
2 | 1 | 24 | 0 | 1 | 5 | 77% | |
3 | 1 | 0 | 15 | 1 | 0 | 88% | |
4 | 0 | 2 | 0 | 22 | 0 | 92% | |
5 | 3 | 0 | 0 | 0 | 8 | 73% | |
Total | – | – | – | – | – | – | 82.1% |
Subtypes: 1, luminal A; 2, luminal BHER2-; 3, luminal BHer2+; 4, HER2 positive; 5, triple negative