Table 6 Four-classes analysis. Model Trained on the 18 features selected in leave 200 out cross validation over 1000 iterations: Confusion matrix.

From: Machine Learning (ML) based-method applied in recurrent pregnancy loss (RPL) patients diagnostic work-up: a potential innovation in common clinical practice

 

Predicted class

0 abortion

1 abortion

2–3 abortions

≥4 abortions

Real class

0 abortion

91.28 ± 0.34%

8.06 ± 0.21%

0.66 ± 0.27%

0.003 ± 0.02%

1 abortion

0.91 ± 0.24%

88.97 ± 0.95%

8.15 ± 0.81%

1.96 ± 0.41%

2–3 abortions

4.16 ± 0.26%

12.98 ± 0.45%

70.12 ± 0.72%

12.75 ± 0.63%

≥4 abortions

2.63 ± 0.17%

9.94 ± 0.41%

11.09 ± 0.89%

76.35 ± 0.94%

  1. ACCub (unbalanced accuracy) = 81.71 ± 0.37%;
  2. ACCb (balanced accuracy) = 81.55 ± 0.37%.
Back to article page