Skip to main content
. 2014 Oct 24;9(10):e108396. doi: 10.1371/journal.pone.0108396

Table 9. Performance of different feature combinations using automatically predicted medical concepts.

Explicit recognizer Implicit recognizer Micro INC
No P R F P R F F
Baseline 0.8123 0.7043 0.7545 0.8217 0.6282 0.7120 0.7443 /
DF1 0.8105 0.7452 0.7765 0.8233 0.6427 0.7219 0.7636 0.0193
DF1+DF2 0.8116 0.7582 0.7840 0.8304 0.6491 0.7286 0.7710 0.0267
DF1+DF2+CF 0.8256 0.7599 0.7914 0.8362 0.6578 0.7363 0.7784 0.0341
DF1+DF2+CF+WF1 0.8495 0.7781 0.8122 0.8452 0.6687 0.7467 0.7967 0.0524
DF1+DF2+CF+WF1+WF2 0.8487 0.7826 0.8143 0.8501 0.6742 0.7200 0.7992 0.0549
DF1+DF2+CF+WF1+WF2+WF3 0.8494 0.7854 0.8160 0.8517 0.6749 0.7531 0.8012 0.0569
DF1+DF2+CF+WF1+WF2+WF3+HF 0.8509 0.7796 0.8137 0.8695 0.6893 0.7690 0.8031 0.0588

*Inc is increment compared with baseline.