A variable normalisée is a statistical term utilisée en analyse de données to refer to a variable that has been adjusted to a common scale. This process is essential when comparing or aggregating data from different sources or with varying units of measurement. Normalization is particularly important in apprentissage automatique and statistics, as it ensures that each variable contributes equally to the analysis, preventing any single variable from disproportionately influencing the results.
La normalisation peut être effectuée en utilisant diverses techniques, notamment :
- Mise à l’échelle min-max: This method rescales the variable to fit within a specified range, typically between 0 and 1. The formula used is:
normalized_value = (x - min(x)) / (max(x) - min(x)). - Normalisation Z-score : Also known as standardization, this technique transforms the data into a distribution with a mean of 0 and a standard deviation of 1. The formula is:
normalized_value = (x - mean) / standard_deviation. - Échelle décimale : This method shifts the decimal point of values, making them smaller while maintaining their relationships.
By normalizing variables, analysts can mitigate issues related to different scales and units, enabling more accurate comparisons and enhancing the performance of machine learning algorithms. For instance, in a dataset containing variables such as income (in dollars) and age (in years), normalization ensures that both variables contribute equally to the model’s predictions.