KI-Glossar
Das vollständige Wörterbuch der Künstlichen Intelligenz
Balanced Random Forest
A variant of Random Forest designed to handle imbalanced datasets by creating decision trees on bootstrap subsamples where each class is equally represented.
Balanced Bootstrap
A sampling technique where, for each iteration, a bootstrap sample is drawn to guarantee equal representation of classes, often by undersampling the majority class or oversampling the minority class.
Majority Class Undersampling
A method to reduce class imbalance by randomly removing observations from the majority class to reduce its predominance in the training dataset.
Minority Class Oversampling
A technique aimed at increasing the number of observations of the minority class, either by duplication or by generating new synthetic observations, to balance the class distribution.
Bootstrap Sample
A random sample drawn with replacement from the original dataset, used in bagging methods to train each model in the ensemble on a slightly different subset of the data.
Gini Score
A measure of node impurity in a decision tree, quantifying the probability that a randomly chosen observation from the node would be misclassified if it were randomly labeled according to the class distribution.
AUC-ROC (Area Under the Receiver Operating Characteristic Curve)
A performance metric that measures a classifier's ability to distinguish between classes, representing the area under the curve that plots the true positive rate against the false positive rate.
EasyEnsemble
An ensemble learning algorithm that creates multiple subsets of the majority class, trains a classifier on each subset combined with the entire minority class, and aggregates the predictions.
BalanceCascade
Iterative ensemble method that sequentially trains classifiers on increasingly balanced datasets, removing correctly classified examples from the majority class at each step.
Recall (Sensitivity)
Metric that measures the proportion of actual positive observations that were correctly identified by the model, essential for evaluating performance on the minority class.