Table 4.
Classification performance obtained with different input feature spaces.
Input features | Accuracy | Precision | F1 score |
---|---|---|---|
Clustered abundances | 72.4 ± 2.1 | 72.7 ± 2.2 | 73.2 ± 2.4 |
10D PCA | 73.3 ± 2.3 | 73.6 ± 2.4 | 73.6 ± 2.4 |
10D MTL | 73.8 ± 2.4 | 74.0 ± 2.5 | 73.8 ± 2.4 |
9D PCA without V | 74.1 ± 2.3 | 74.2 ± 2.5 | 74.1 ± 2.4 |
Abundances without Co | 75.0 ± 2.1 | 75.3 ± 1.9 | 75.4 ± 2.3 |
All abundances | 75.2 ± 2.3 | 75.5 ± 2.3 | 75.7 ± 2.3 |
Abundances without V, Co, TiII | 76.0 ± 2.3 | 76.5 ± 2.4 | 75.5 ± 2.2 |
Abundances without V, Co | 76.1 ± 2.2 | 76.8 ± 2.3 | 75.7 ± 2.2 |
Abundances without V, TiII | 76.2 ± 2.2 | 76.6 ± 2.3 | 76.0 ± 2.4 |
Abundances without V | 76.4 ± 2.4 | 76.7 ± 2.5 | 76.5 ± 2.4 |
Notes. Classification performance (expressed as percentages), averaged across folds, obtained with different input feature spaces, using only high-quality clusters and a RBF SVM classifier. Clustered abundances refers to the subset of chemical elements defined through correlation clustering: Si, C, Mg, Fe, Ti, S, Na, N, V, TiII (see Fig. 3).
Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.
Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.
Initial download of the metrics may take a while.