KI-Glossar
Das vollständige Wörterbuch der Künstlichen Intelligenz
Random Feature Selection
Technique introduced in Random Forest where each decision tree is trained on a random subset of features at each split, reducing correlation between trees and improving generalization.
Feature Importance
Metric calculated in Random Forest that evaluates the contribution of each variable to model accuracy, typically measured by the mean decrease in impurity (Gini or entropy) or the increase in OOB error when the variable is permuted.
Majority Voting Aggregation
Aggregation method for classification problems where the final prediction of the ensemble is the class that receives the most votes among the individual predictions of the base models.
Extremely Randomized Trees
Variant of Random Forest that introduces more randomization by choosing feature split thresholds completely randomly rather than searching for the optimal threshold, reducing bias and variance.
Proximity Between Observations
Similarity matrix calculated in Random Forest that counts the number of times two observations end up in the same terminal leaf across all trees in the forest, which can be used for anomaly detection or clustering.