CAJ | 학술논문

为了改善单个分类器对雷达目标的识别性能，分别从差异性和准确率两个因素出发，提出了一种通过变换分类器选择集成对雷达目标识别的算法。首先，为了增加集成个体差异性，将个体分类器的预测标记作为变换的原始目标，将正确标记引入来代替变换中的目标均值，以此构造一个整数矩阵。通过将个体分类器的预测标记投影到经过正确标记的直线上，从而获得一组新的预测标记。然后，根据准确率和 RPF-measure这两种衡量分类器性能的准则，选择一些性能提高了的个体进行集成，以此保证个体分类器的准确率。最后，通过结合被选择的个体预测标记来改善对雷达目标的识别性能。对于UCI机器学习资源库中数据集和雷达一维距离像的实验结果表明，该算法能有效平衡个体差异性和个体准确率两个因素，并且相比单个分类器和其他集成方法，该方法提高了对雷达目标的识别准确率。
위료개선단개분류기대뢰체목표적식별성능，분별종차이성화준학솔량개인소출발，제출료일충통과변환분류기선택집성대뢰체목표식별적산법。수선，위료증가집성개체차이성，장개체분류기적예측표기작위변환적원시목표，장정학표기인입래대체변환중적목표균치，이차구조일개정수구진。통과장개체분류기적예측표기투영도경과정학표기적직선상，종이획득일조신적예측표기。연후，근거준학솔화 RPF-measure저량충형량분류기성능적준칙，선택일사성능제고료적개체진행집성，이차보증개체분류기적준학솔。최후，통과결합피선택적개체예측표기래개선대뢰체목표적식별성능。대우UCI궤기학습자원고중수거집화뢰체일유거리상적실험결과표명，해산법능유효평형개체차이성화개체준학솔량개인소，병차상비단개분류기화기타집성방법，해방법제고료대뢰체목표적식별준학솔。
Diversity among individuals and accuracy of individuals are two important factors to decide the ensemble generalization error,whereas enhancing diversity is at the cost of decreasing the accuracy of individuals.Hence,in order to improve the performance of radar target recognition classified by a single classifier,this paper introduces a new radar target recognition method based on the integer matrix linear transformation selective classifier ensemble that considers the balance of diversity and accuracy.Firstly,in order to enhance the diversity,the individual classifiers are considered as original targets of the linear transformation,and instead of the mean value of samples,the true labels are considered to construct an integer matrix.By projecting individual classifiers on the lines through the true labels,a set of new classifiers is obtained based on the proj ect transformation.Secondly,according to two rules that measuring the performance of the classifier,the accuracy rate and RPF-measure,some new classifiers that can obtain better performance are selected to ensemble for increasing the accuracy of classifiers of an ensemble. Finally,the performance of radar target recognition is improved by combining the selected new classifiers. Experimental results of UCI datasets and the radar range profile indicate that the proposed method balances effectively diversity and accuracy,and that it can obtain better performance for radar target recognition compared with single classifier algorithms and other methods.