An アウト・オブ・バッグサンプル (OOB sample) refers to a collection of data points that are not included in the training set when creating a predictive model, particularly in アンサンブル手法 like bagging. In these methods, multiple models are trained on random subsets of the data, and the points that are not selected for a specific model’s training are considered its アウト・オブ・バッグサンプル。
These OOB samples can serve a dual purpose: they provide a means of estimating the model’s performance without needing a separate validation set, and they help in evaluating the generalization ability of the model. For instance, in a ランダムフォレスト algorithm, each decision tree is built using a bootstrap sample drawn from the training data. Approximately one-third of the data points are left out of the bootstrap sample and thus are available for OOB validation.
評価するために モデルのパフォーマンス using OOB samples, predictions can be made for each out-of-bag observation using the trees in the forest that did not include that observation in their training set. The aggregated results from these predictions allow for an unbiased estimate of the model’s accuracy.
Using out-of-bag samples is particularly advantageous because it maximizes data utility by leveraging the same dataset for both training and validation, thereby avoiding the pitfalls of overfitting and underfitting 通常、別々のトレーニングと検証データセットに関連付けられています。