Table 6.
Cut-off threshold | Predicted interologs | True positives | Putatives | Precision | Recall |
CS ≥ 0 | 90,871 | 2,572 | 88,299 | 2.83% | -* |
CS ≥ 1 | 59,919 | 2,473 | 57,446 | 4.13% | 96.15% |
CS ≥ 2 | 41,828 | 2,222 | 39,606 | 5.31% | 86.39% |
CS ≥ 3 | 27,048 | 1,772 | 25,276 | 6.55% | 68.90% |
CS ≥ 4 | 19,659 | 1,467 | 18,192 | 7.46% | 57.04% |
CS ≥ 5 | 14,344 | 1,226 | 13,118 | 8.55% | 47.67% |
CS ≥ 6 | 11,334 | 1,021 | 10,313 | 9.01% | 39.70% |
CS ≥ 7 | 8,374 | 859 | 7,515 | 10.26% | 33.40% |
* : the predicted data set do not contain negative data to calculate false negative (FN) to obtain recall = (TP/(TP+FN))