基于贝叶斯和F-measure感知机的分类器设计  

Design of classifier based on Bayes and F-measure perceptron

作　　者：马占杰 杨淑莹[2] MA Zhanjie;YANG Shuying(Department of Computer Science and Engineering,Tianjin University of Technology,Tianjin 300384,China;Key Laboratory of Computer Vision and Systems,Ministry of Education,Tianjin University of Technology,Tianjin 300384,China)

机构地区：[1]天津理工大学计算机科学与工程学院,天津300384 [2]计算机视觉与系统教育部重点实验室,天津300384

出　　处：《现代电子技术》2019年第21期125-129,共5页Modern Electronics Technique

基　　金：国家自然科学基金(61001174);天津市科技支撑计划;天津市自然科学基金(13JCYBJC17700)~~

摘　　要：当样本特征向量交织时,分类很容易出错。为解决这个问题,提出一种基于Bayes和F-measure的分类器算法。采用替代方法评估分类器的性能正受到关注,特别是对于不平衡的问题。该算法利用F-measure分析不平衡数据的分类准确度,将类概率密度函数引入判据,并采用梯度下降法得到准则函数。文中将所提出的方法与传统方法进行比较,实验结果表明,该方法能够有效提高识别的准确率和精确度。A classifier algorithm based on Bayes and F-measure is proposed to solve the problem that the classification is prone to error when the sample feature vectors are intertwined.The alternative methods used for the performance evaluation of classifiers are receiving increasing attention,especially for unbalanced data classification.The algorithm is used to analyze the classification accuracy of the unbalanced data by means of F-measure.The probability density function is introduced into the criterion,and the gradient descent method is used to obtain criterion function.The proposed method is compared with the traditional ones,in this paper.The experimental results show that the proposed method can effectively improve the accuracy and precision of recognition.

关 键 词：分类 F-MEASURE 不平衡数据 后验概率 准确率 实验验证 

分 类 号：TN02[电子电信—物理电子学] 34