Dynamic Classifier Selection with Confidence Intervals

Nowadays, the ensembles are a popular classification method. In order to obtain the final decision the selection and the fusion methods are used. In this paper, the Dynamic Classifier Selection with Confidence Intervals (DCS-CONFI) method is proposed. This method use confidence intervals for identity the true knowledge or the influence of each individual classifier in the final decision, thus, the member with higher confidence interval is chosen tor classify the test pattern. The experimental results demonstrated the convenience of to determinate the confidence level when the classifier selection scheme is used.