Table 2.

Constant and Variable Hyperparameters for Each Machine-Learning Algorithm

Algorithm	Hyperparameters
	Constant Values	Values Tested
SGD	Loss: Logistic regression Penalty: ElasticNet	Learning rate: 10-5–10-2 Epochs: 5–1,000 by 5
SVC	Kernel: Radial basis function	Penalty C term: 1, 10, 100 Gamma: 10-5–10-1
Random forest		Estimators: 10–190 by 10
DNN	Early stopping: No improvement in loss function for 30 epochs Learning rate optimizer: Adam Loss: Binary cross entropy Neuron activation function: ReLu Output activation function: Sigmoid	Neurons: 23–26 Hidden layers: 0, 1, 2, 4, 6

Note: SGD = stochastic gradient descent; SVC = support vector classifier; DNN = dense neural network