Skip to main content

Table 3 Logistic Regression Results: One-Versus-One Classification. ROC-AUC scores for one-versus-one logistic regression experiments for all category combinations. Feature sets consist of the number of revisions (nrev), control variables (CV), network metrics (NM), label frequencies (LF), and label transition probabilities (LT). Utilizing all feature sets performs best for all experiments

From: Relating Wikipedia article quality to edit behavior and link structure

Features ROC-AUC
  LQ vs HQ LQ vs CI LQ vs EW HQ vs CI HQ vs EW CI vs EW
nrev 0.577 0.870 0.968 0.849 0.961 0.810
CV 0.652 0.923 0.989 0.883 0.978 0.837
NM 0.707 0.861 0.938 0.871 0.937 0.710
LF 0.640 0.807 0.880 0.881 0.928 0.665
LT 0.713 0.825 0.907 0.884 0.933 0.740
CV + NM 0.727 0.930 0.987 0.919 0.978 0.833
CV + LF 0.677 0.947 0.992 0.951 0.991 0.850
CV + LT 0.736 0.952 0.994 0.947 0.992 0.874
NM + LF 0.742 0.906 0.967 0.950 0.980 0.738
NM + LT 0.775 0.923 0.974 0.949 0.978 0.788
LF + LT 0.744 0.892 0.956 0.953 0.980 0.769
NM + LF + LT 0.799 0.947 0.986 0.979 0.993 0.813
CV + NM + LF 0.755 0.954 0.991 0.967 0.992 0.840
CV + NM + LT 0.783 0.959 0.994 0.963 0.993 0.863
CV + NM + LF + LT 0.805 0.973 0.996 0.984 0.997 0.879