Estimating the Accuracy of a Bagged Ensemble

In ensemble machine learning, we combine the decisions of weak learners to derive a decision that is, hopefully, better than the individual ones. The combination of these learners can be aggregated by a majority vote or simple averaging, or it can be more complicated and involve multiple steps such as in boosting. In this paper, we consider the question of predicting the accuracy of an ensemble created with bagging for a given number of weak learners. We achieve a low relative error on our predictions and can make this prediction in a shorter time, as compared to training the ensemble over various sizes.