Table 4.
Performance of linear discriminant analysis models by recursive elimination of descriptors.
Top Descriptors | Full Model | LOO-CV | Delta |
---|---|---|---|
12 | 53/70 (0.75714) | 36/70 (0.51429) | 0.24286 |
11 | 52/70 (0.74286) | 36/70 (0.51429) | 0.22857 |
10 | 50/70 (0.71429) | 35/70 (0.5) | 0.21429 |
9 | 50/70 (0.71429) | 38/70 (0.54286) | 0.17143 |
8a | 49/70 (0.7) | 39/70 (0.55714) | 0.14286 |
7 | 48/70 (0.6851) | 39/70 (0.55714) | 0.12857 |
6 | 46/70 (0.65714) | 34/70 (0.48571) | 0.17143 |
5 | 44/70 (0.62857) | 33/70 (0.47143) | 0.15714 |
4 | 45/70 (0.64286) | 36/70 (0.51429) | 0.12857 |
3 | 43/70 (0.61429) | 37/70 (0.52857) | 0.085714 |
2 | 34/70 (0.48571) | 31/70 (0.44286) | 0.042857 |
The model utilizing the top eight descriptors is considered to be the optimal model with its overall performance and predictability performance relative to the other 10 models.