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

Table 12 Mean XGBoost accuracies and F1-scores (in %) for the AIBL data set (no information rate 57.14%)

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	55.00	57.97	444	67.50	73.44
LOO (LR)	84	59.64	66.21	69	63.21	70.00	422	68.21	73.68
LOO (RF)	49	57.14	64.69	11	61.07	68.43	97	66.43	73.53
Data Shapley (LR)	152	63.21	69.97	248	58.57	66.66	396	75.36	77.53
Data Shapley (RF)	164	62.14	71.65	133	61.79	70.84	195	66.43	75.34

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

ISSN: 1758-9193