Ranking

Legende

Unter Quellcode sind die Präfixe zu den KNIME Workflows und Anacona (Python) Juypter Notebooks zu finden. Mit Version ist die jeweilige Datenversion gemeint. Der Mittelwert berechnet sich zu gleichen Anteilen aus Genauigkeit (Accuracy), Präzision (Precision), F1-Maß (F1-Score) und AURC [Area Under ROC Curve, ROC = Receiver Operating Characteristic]. Die letzten vier Spalten enthalten die Einträge der Konfusionsmatrix (Confusion Matrix): TN [True Negatives], FP [False Positives], FN [False Negatives], TP [True Positives].

Methode

m = Manual Model, knn = k-Nearest Neighbor, nb = Naive Bayes, dt = Decision Tree, mlp = Multilayer Perceptron, svm = Support Vector Machines, lr = Logistic Regression, rf = Random Forest, gb = Gradient Boosting, gbt = Gradient Boosted Trees, te = Tree Ensemble.

Parameter

opt. = Optimized Model, minmax = Min-Max Scaler, standard = Standard Scaler, poly = Polynomial Function, sag = Stochastic Average Gradient, laplace = Laplace Distribution, uniform = Uniform Distribution, rbf = Radial Bias Function, linear = Linear Function, newton-cg = Newton Conjugate Gradients, ffs = Forward Feature Selection, bfe = Backward Feature Elimination, random = Random Search, hillclimbing = Hillclimbing Search, bayesian = Bayesian Search, gs = Grid Search, rs = Randomized Search, hgs = Halving Grid Search, hrs = Halving Random Seach, thr. = Threshold, bf = Brute Force, acc. = Accuracy, pre. = Precision, miss val = Missing Values.