Explainability of random survival forests in predicting conversion risk from mild cognitive impairment to Alzheimer’s disease

Brain Informatics

Table 2 Hyperparameters of Cox proportional hazard (CPH) and Random Survival Forests (RSF)

	Hyperparameter	Parameter distribution	Optimal value
CPH	l2_reg	float from a reciprocal continuous random distribution in range (0.1, 100)	99.4
CPH	lr	float from a reciprocal continuous random distribution in range (0.1, 1)	0.79
RSF	importance_mode	['normalized_permutation','permutation','impurity','impurity_corrected']	‘permutation'
	max_depth	integer from a reciprocal continuous random distribution in range (5, 50)	24
	min_node_size	integer from a reciprocal continuous random distribution in range (1, 40)	7
	max_features	[‘all’,'sqrt','log2']	‘sqrt’
	sample_size_pct	[0.60,0.70,0.80,0.90]	0.80

Tuning was performed through a randomized search with threefold cross-validation and 50 repetitions