Skip to main content

Table 3 Comparison of prediction performance of different classification models

From: Time-to-event prediction analysis of patients with chronic heart failure comorbid with atrial fibrillation: a LightGBM model

  Logistic regression LightGBM
1-year 2-year 3-year 1-year 2-year 3-year
AUC 0.687
(0.680,0.694)
0.667
(0.660, 0.673)
0.683
(0.677, 0.689)
0.718*
(0.710, 0.727)
0.744*
(0.737, 0.751)
0.757*
(0.751, 0.763)
Accuracy 0.709
(0.705, 0.713)
0.715
(0.710, 0.720)
0.732
(0.727, 0.737)
0.853
(0.850, 0.857)
0.855
(0.852, 0.859)
0.864
(0.861, 0.868)
Sensitivity 0.662
(0.646, 0.677)
0.606
(0.593, 0.619)
0.618
(0.605, 0.631)
0.559
(0.542, 0.576)
0.603
(0.589, 0.617)
0.615
(0.603, 0.626)
Specificity 0.713
(0.708, 0.717)
0.727
(0.722, 0.734)
0.748
(0.742, 0.754)
0.878
(0..875, 0.881)
0.885
(0.882, 0.888)
0.900
(0.897, 0.903)
f-measure 0.257
(0.250, 0.264)
0.308
(0.300, 0.315)
0.363
(0.355, 0.371)
0.367
(0.356, 0.378)
0.465
(0.455,0.475)
0.528
(0.519, 0.537)
Brier score 0.291
(0.287, 0.295)
0.285
(0.280, 0.290)
0.268
(0.263,0.273)
0.146
(0.143, 0.150)
0.145
(0.141, 0.148)
0.135
(0.132, 0.138)
  1. Data are presented as mean (95% confidence interval), CI: confidence interval, AUC: area under the curve
  2. *DeLong test, P < 0.05, the AUC of 1-, 2-, and 3-year all-cause mortality were different between LightGBM and logistic regression model