Hors-échantillon Évaluation refers to the process of assessing the performance of an intelligence artificielle (AI) model using data that was not included during the model’s training phase. This evaluation is crucial for understanding how well the model can generalize its learned patterns to new, unseen data, which is a key indicator of its effectiveness in real-world applications.
En IA et apprentissage automatique, models are trained on a specific dataset, known as the training set. However, if we only evaluate the model on this training set, we may obtain an overly optimistic view of its performance. This is because the model may simply memorize the données d'entraînement instead of learning to generalize. To combat this issue, out-of-sample evaluation is performed using a separate dataset, often called the test set or validation set, which contains data that the model has not encountered before.
Les techniques courantes pour réaliser des évaluations hors échantillon incluent :
- Méthode de réserve : Splitting the entire dataset into a training set and a test set. The model is trained on the training set and evaluated on the test set.
- Validation croisée K-fold : Dividing the dataset into ‘k’ subsets. The model is trained ‘k’ times, each time using a different subset as the test set, while the remaining subsets are used for training. This method provides a more robust evaluation.
- Validation croisée Leave-One-Out : A special case of k-fold cross-validation where ‘k’ is equal to the number of instances in the dataset. Each instance is used once as a test set while the remaining instances form the training set.
Dans l’ensemble, l’évaluation hors-échantillon est une étape fondamentale dans le développement de modèles lifecycle, ensuring that the AI system is reliable and effective in practical scenarios.