Exploration of classification confidence in ensemble learning

Background Check: A General Technique to Build More Reliable and Versatile Classifiers

“…The benefits of considering the confidence of a classifier's predictions have been discussed before in the machine learning literature [16]- [19]. Most approaches treat a classifier's outputs or class-conditional probability estimates as confidence.…”

“…Most approaches treat a classifier's outputs or class-conditional probability estimates as confidence. [19] discussed the importance of confidence in weighted voting schemes in ensemble learning. Their approach associates a confidence level to the prediction given by each classifier of the ensemble for a given example.…”

“…In the particular application of ensemble learning proposed by [19] every classifier of the ensemble is built on a subset of the training data. Hence, even if no new classes are expected during deployment, these models should not be overconfident about instances that come from regions of the feature space on which they were not trained.…”

The Performance of Individual and Ensemble Classifiers for an Arabic Sign Language Recognition System

“…Then, one makes a final prediction using all the predictions of the other algorithms (re-sampling) as inputs. The re-sampling method can be one of the following: sum, maximum, minimum, and weighted majority voting of the predictions that the other algorithms have generated as extra inputs [17].…”

“…They attempted several strategies to use the classification confidences and the weights. They observed that weighted voting based on classification confidence is better than simple voting if all of the base classifiers are used [17].…”