Table 4 The best-tuned hyperparameters for each model

Gradient Boosting

max_depth = 10, max_features = 'sqrt', min_samples_split = 50, n_estimators = 800, random_state = 8, learning_rate = 0.5, subsample = 0.5

Random Forests

max_depth = 60, max_features = 'sqrt', min_samples_split = 5, min_samples_leaf = 4, n_estimators = 400, random_state = 8

Artificial Neural Network

activation = 'identity', alpha = 0.0001, batch_size = 'auto', hidden_layer_sizes = 7, learning_rate = 'adaptive', learning_rate_init = 0.001, max_iter = 500, solver = 'lbfgs'

Logistic Regression

C = 0.4, multi_class = 'multinomial', random_state = 8, solver = 'saga'

Naive Bayes

alpha = 1.0, fit_prior = True, class_prior = None

Support Vector Machine

C = 0.1, degree = 4, kernel = 'poly', probability = True, random_state = 8

K-Nearest Neighbors

n_neighbors = 3