Skip to main content

Table 4 The χ2 p-values for the fit to the diagonal in the reliability diagram, number of calibrated points, and difference between the maximum and minimum calibrated probabilities (range) for various BBQ parameters Fig. 3

From: A nonparametric Bayesian method of translating machine learning scores to probabilities in clinical decision support

Classifier Scoring function Threshold (α) Binning parameter (N 0) χ 2 p-value Calibration points Range
SVM BDeu 2 0.0001 0.187 7 0.95
SVM BDeu 4 0.0001 0.13 8 0.97
SVM BDeu2 N/A 0.0001 0.023 9 0.97
SVM BDeu 2 0.001 0.187 7 0.95
SVM BDeu 4 0.001 0.13 8 0.97
SVM BDeu2 N/A 0.001 0.048 9 0.94
SVM BDeu 2 0.01 0.187 7 0.95
SVM BDeu 4 0.01 0.13 8 0.97
SVM BDeu2 N/A 0.01 0.066 9 0.94
k-means BDeu 2 0.0001 0.502 2 0.05
k-means BDeu 4 0.0001 0.558 2 0.06
k-means BDeu2 N/A 0.0001 0.497 2 0.05
k-means BDeu 2 0.001 0.502 2 0.05
k-means BDeu 4 0.001 0.558 2 0.06
k-means BDeu2 N/A 0.001 0.497 2 0.05
k-means BDeu 2 0.01 0.502 2 0.05
k-means BDeu 4 0.01 0.558 2 0.06
k-means BDeu2 N/A 0.01 0.496 2 0.05
  1. The BBQ default parameters used in the comparisons above are highlighted in boldface