Table 4 illustrates the optimal-performing MLM and preprocessing, as well as tuned hyper-parameters with the highest AAC
From: Prediction of diabetes disease using an ensemble of machine learning multi-classifier models
MLMs | Optimal-performing preprocessing | Hyper-parameter tuning methods | Optimal hyper-parameters | Performance |
|---|---|---|---|---|
k-NN | I + N MRMR = 6 | Grid search | Algorithm = auto leaf_size = 5 n_neighbors = 25 weight = uniform \({L}_{2}\)- norm (Euclidean) | \(0.971\pm 0.003\) |
SVM | I + N | Bayesian optimization | C = 1 Gamma = 0.1 Kernel = RBF Decision_function_shape = OVO | \(0.948\pm 0.003\) |
DT | I + Z MRMR = 10 | Bayesian optimization | Criterion = gini bootstrap = True min_samples_leaf = 1 max_depth = 8 max_features = auto min_samples_leaf = 2 min_samples_split = 0.2 | \(0.968\pm 0.003\) |
RF | I + N MRMR = 6,8,10 | Bayesian optimization | Criterion = gini n_estimator = 150 bootstrap = True min_samples_leaf = 1 max_depth = 8 max_features = sqrt | \(0.988\pm 0.003\) |
GNB | I + Z + PCA = 12 | Grid search | var_smoothing = 08112 | \(0.926\pm 0.006\) |
AdaBoost | I + MMR = 10 | Grid search | boosting algorithm = AdaBoost.MH n_estimator = 150 learninh_rate = 0.1 | \(0.961\pm 0.003\) |