From: Machine learning based analyses on metabolic networks supports high-throughput knockout screens
Machine Learning | Machine Learning | |
---|---|---|
(all 30 features) | (25 optimized features) | |
true positives | 168 | 174 |
true negatives | 1078 | 1092 |
false positives | 47 | 33 |
false negatives | 63 | 57 |
sensitivity (recall) | 72.73% | 75.32% |
specificity | 95.82% | 97.07% |
positive predictive values (precision) | 78.14% | 84.06% |
negative predictive values | 94.48% | 95.04% |
overall accuracy | 91.89% | 93.36% |