F1Werte

F1Werte, in German often written F1-Werte, denote a family of performance metrics used to evaluate classification models, centered on the F1 score, the harmonic mean of precision and recall. Precision measures the accuracy of positive predictions, while recall assesses the ability to identify actual positives. The standard F1 score is defined as F1 = 2 × precision × recall / (precision + recall).

In practice, F1Werte can be computed in several forms for different evaluation needs. The macro-F1 averages the F1 scores across classes, treating each class equally. The micro-F1 aggregates true positives, false positives, and false negatives across classes before computing the score, reflecting overall performance. The weighted-F1 combines per-class F1 scores weighted by class support, giving more influence to common classes. An extended family includes Fβ scores, where β controls the relative importance of recall versus precision; F1 corresponds to β = 1.

F1Werte are widely used in information retrieval, natural language processing, computer vision, and medical diagnostics. They are particularly helpful when there is a need to balance precision and recall, or when class distribution is imbalanced. However, they have limitations: F1Werte do not capture calibration of probabilities, can obscure differences between models that trade precision and recall differently, and can be sensitive to class prevalence in multi-class settings.

Overall, F1Werte provide a concise, interpretable single-number summary of a model’s ability to identify positive instances while maintaining reasonable precision, making them a standard tool in model evaluation alongside other metrics such as precision, recall, and area under the ROC curve.