Modellverschiebung is a phenomenon in maschinellem Lernen where the performance of a model declines over time as the statistical properties of the input data change. This can happen for various reasons, such as shifts in the underlying Datenverteilung, changes in user behavior, or the introduction of neue Daten Quellen.
In practical terms, when a model is trained, it learns to make predictions based on the patterns present in the Trainingsdaten. If the data it encounters during deployment differs significantly from this training data—due to factors like seasonality, market trends, or evolving user preferences—the model may no longer perform as expected. This leads to a situation where the model’s predictions become less accurate, which is detrimental in applications ranging from financial forecasting to healthcare diagnostics.
Es gibt zwei Haupttypen von Modellverschiebung: Kovariatenverschiebung and Konzept-Drift. Covariate shift occurs when the distribution of the input features changes, while concept drift involves changes in the relationship between the input features and the target variable. Both types of drift can significantly impact the model’s effectiveness.
Um den Modell-Drift zu steuern, setzen Praktiker oft um Überwachungssysteme that track model performance over time. When a decline in accuracy is detected, they may choose to retrain the model on more recent data, adjust the model architecture, or employ techniques such as ensemble learning to maintain performance. Regularly updating models helps ensure that they remain relevant and effective in dynamic environments.