Log-Verlust, also known as logistic loss or Kreuzentropie-Verlust, is a performance metric used primarily in binärer Klassifikation problems where the predicted output is a probability value ranging between 0 and 1. It quantifies the difference between the predicted probabilities and the actual class labels (0 or 1). The objective of using log loss is to evaluate how well a classification model predicts probabilities for binary outcomes, with lower values indicating better Modellleistung.
Mathematisch wird der Log-Verlust mit der folgenden Formel berechnet:
Log Loss = -1/N * Σ [y * log(p) + (1 – y) * log(1 – p)]
Wo:
- N ist die Gesamtzahl der Vorhersagen.
- y ist das tatsächliche Label (0 oder 1).
- p ist die vorhergesagte Wahrscheinlichkeit der positiven Klasse (1).
The Log Loss value ranges from 0 to infinity, where 0 indicates perfect predictions (model outputs probabilities of either 0 or 1 correctly) and larger values indicate worse performance. A model that predicts probabilities close to the true labels will have a lower log loss, while a model that predicts probabilities far von den wahren Labels abweicht, werden einen höheren Log Loss verursachen.
Log Loss is particularly useful in scenarios where the output is not just a hard classification but a probability, making it suitable for applications such as logistische Regression, neural networks, and other probabilistic classifiers. It is also widely im maschinellen Lernen Wettbewerbe, wie sie von Kaggle veranstaltet werden, um die Modellleistung zu bewerten.