Table 4.

Implementation parameters for the classification methods

Method	Implementation
Decision trees	• Splitting criterion: Gini index
	• Pre-pruning: minimum of 1, 3 and 5 instances in lead nodes
	• Post-pruning: test all admissible prune levels between minimum and maximum values for each tree
Naïve Bayes	• Feature distributions: multivariate multinomial (discrete), kernel estimation (continuous)
	• Classification threshold: from 0 to 1 in steps of 0.005
Logistic regression	• Classification threshold: from 0 to 1 in steps of 0.005
Support Vector Machines	• Linear kernels
	• Penalty parameter (C) ${10}^{-2}$ to ${10}^2$ (unitary exponent increments)
Deep Neural Network	• Stochastic Gradient Descent
	• Number of epochs: 10
	• Softmax activation function
	• Learning rate ${10}^{-1}$ to ${10}^{-6}$ (unitary exponent increments)