La mise à l'échelle Min-Max, également connue sous le nom de Normalisation Min-Max, is a technique de prétraitement des données used to transform features to a specific range, usually between 0 and 1. This method is particularly useful in apprentissage automatique and statistics où les algorithmes sont sensibles à l'échelle des caractéristiques d'entrée.
La formule pour la mise à l'échelle Min-Max est :
X' = (X - X_{min}) / (X_{max} - X_{min})où :
X'is the scaled value.Xis the original value.X_{min}is the minimum value of the feature in the dataset.X_{max}is the maximum value of the feature in the dataset.
This technique is beneficial because it preserves the relationships between the data points while scaling them to a uniform range. It ensures that each feature contributes equally to the distance calculations in algorithms such as k-plus proches voisins (KNN) ou méthodes de descente de gradient.
However, Min-Max Scaling has its limitations. It is sensitive to outliers, which can significantly skew the scaled values if the minimum or maximum values are extreme. In such cases, alternative normalization methods, like Z-score normalization, may be more appropriate. Overall, Min-Max Scaling is a straightforward and effective method for mise à l'échelle des caractéristiques dans diverses applications de la science des données.