Skip to main content

Table 10 Mean XGBoost accuracies and F1-scores (in %) for the independent ADNI test set (no information rate 55.56%)

From: Data analysis with Shapley values for automatic subject selection in Alzheimer’s disease data sets using interpretable machine learning

Exclusion method Zero-cutoff Max valid Max test
  n ACC (in %) F1 (in %) n ACC (in %) F1 (in %) n ACC (in %) F1 (in %)
Random (-) 430 62.78 64.45 391 66.39 68.40
LOO (LR) 84 60.56 63.50 69 60.62 65.05 178 63.57 65.28
LOO (RF) 49 61.25 65.84 11 59.38 63.96 407 64.03 68.21
Data Shapley (LR) 152 68.06 71.02 248 65.07 67.89 144 68.47 71.22
Data Shapley (RF) 164 66.88 72.12 133 65.90 70.62 178 68.47 73.48
  1. Different methods were used to identify and focus on the training subjects with the most informative data. The zero-cutoff method excluded all training subjects with Data Shapley values smaller than zero. Max valid was the threshold achieved by maximizing the results for the independent validation set. Max test was the optimistic threshold which achieved the best results for the test set. Ten repetitions with different seeds were performed for every exclusion data set. The best results are highlighted in bold