New Dynamic Ensemble of Classifiers Selection approach based on confusion matrix for Arabic Handwritten recognition

Abstract

Arabic Pattern recognition can be regarded as a problem of classification, where different patterns are presented and be needed to classify into specified classes. One way to improve the recognition rates of pattern recognition tasks is to improve the accuracy of individual classifiers, and another is to apply ensemble of classifiers methods. The advantage of dynamic ensemble selection vs dynamic classifier selection is that we distribute the risk of this over-generalization by choosing a group of classifiers instead of one individual classifier for a test pattern. In this paper we propose a new approach for Arabic handwriting recognition based on dynamic selection of ensembles of classifiers. Our DECS-LA algorithm (Dynamic Ensemble of Classifiers Selection by Local accuracy) chooses the most confident ensemble of classifiers to label each test sample dynamically. Such a level of confidence is measured by calculating the proposed local reliability measure using confusion matrixes constructed during training level. After validation with voting and weighted voting and ten different classifiers, we show experimentally that choosing classifier ensembles dynamically taking into account the behavior classifier during test level by L-Reliability measure leads to increase recognition rate for Arabic Handwritten recognition system.