Skip to main content

Table 3 Predictive performance of models using pre-assigned (P) or learned (L) weights in the weighted sampling (WS) strategy

From: Learning temporal weights of clinical events using variable importance

  Accuracy   AUC   AUPRC
ADE PWS LWS   PWS LWS   PWS LWS
D611 76.46 90.38   0.887 0.946   0.949 0.981
D642 97.79 99.47   0.979 0.983   0.814 0.936
D695 71.10 87.74   0.895 0.923   0.806 0.881
E273 81.78 79.92   0.615 0.650   0.264 0.272
G620 91.64 92.22   0.815 0.837   0.334 0.441
I952 91.71 91.64   0.557 0.554   0.116 0.118
L270 67.14 73.07   0.843 0.828   0.791 0.795
L271 84.18 85.22   0.753 0.759   0.356 0.406
M804 95.17 95.12   0.571 0.598   0.072 0.079
M814 98.71 98.69   0.702 0.722   0.039 0.052
O355 82.13 96.49   0.980 0.989   0.948 0.977
R502 99.16 99.22   0.840 0.774   0.148 0.280
T782 94.81 94.86   0.741 0.754   0.188 0.255
T783 83.33 86.16   0.771 0.788   0.393 0.524
T784 66.86 75.46   0.762 0.761   0.830 0.829
T801 90.35 93.55   0.868 0.880   0.533 0.635
T808 77.93 84.38   0.869 0.879   0.954 0.954
T886 94.95 94.86   0.741 0.738   0.159 0.190
T887 60.19 69.72   0.762 0.771   0.765 0.802
Average 84.49 88.85   0.787 0.797   0.498 0.548
p-value 0.002838   0.01597   0.00001907
  1. Bold indicates winning