Table 4
Results for KDE classification.
| Equal priors | ||
| Absolute | ||
| bhb | other | |
| BHB | 1021 | 222 |
| OTHER | 505 | 713 |
| Percent | ||
| bhb | other | |
| BHB | 82.1 % | 17.9 % |
| OTHER | 41.5 % | 58.5 % |
| Completeness | Contamination | |
| BHB | 0.821 | 0.512 |
|
|
||
| With simple prior | ||
| bhb | other | |
| BHB | 765 | 478 |
| OTHER | 281 | 937 |
| Percent | ||
| BHB | 61.5 % | 38.5 % |
| OTHER | 23.1 % | 76.9 % |
| Completeness | Contamination | |
| BHB | 0.615 | 0.438 |
|
|
||
| With 2d prior | ||
| bhb | other | |
| BHB | 912 | 331 |
| OTHER | 293 | 925 |
| Percent | ||
| BHB | 73.4 % | 26.6 % |
| OTHER | 24.1 % | 75.9 % |
| Completeness | Contamination | |
| BHB | 0.734 | 0.405 |
Notes. Top section: confusion matrix showing results of KDE classification of 10 independently selected test sets based on 10 independently trained models. The classification threshold is P(BHB) ≥ 0.5 in each case. The results are shown at the top as mean numbers of sources in each category and then as percentages. The completeness and contamination for the BHB output sample is also shown. These are corrected for the expected unbalanced class fractions. Middle section: the same quantities calculated with the simple prior applied to the classifier output probabilities. Bottom section: the confusion matrix, BHB completeness and BHB contamination obtained when the prior as a function of g and latitude is applied to the classifier output probabilities.
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.