L'Erreur Absolue est un terme utilisé dans statistics and analyse de données to quantify the accuracy of a prediction or measurement. It is defined as the absolute difference between the predicted value (or valeur observée) et la valeur réelle (ou valeur effective). Mathématiquement, elle peut s'exprimer comme :
Absolute Error = | Predicted Value – Actual Value |
Cette métrique est importante car elle offre une manière simple d'évaluer à quel point far off a prediction is from the actual result, regardless of the direction of the error (whether the prediction is above or below the actual value). As a result, Absolute Error is always a non-negative number.
Dans le contexte de l'IA et apprentissage automatique, understanding Absolute Error helps in evaluating the performance of models. For instance, if you are building a regression model to predict housing prices, the Absolute Error will help you understand how close your model’s predictions are to the actual sale prices of the houses. By calculating the Absolute Error across all predictions, you can derive insights about the overall accuracy of the model and identify areas for improvement.
While Absolute Error is a useful metric, it does not provide a normalized view of error sizes, which is why it is often used in conjunction with other metrics, such as Erreur Absolue Moyenne (MAE) or Root Mean Squared Error (RMSE), which average the Absolute Errors across multiple observations for a more comprehensive assessment of model performance.