基于边界条件GAN的不平衡大数据模糊分类

Fuzzy Classification of Unbalanced Big Data Based on Boundary Condition GAN

针对大数据分类中的不平衡问题,本文提出一种基于边界条件生成式对抗网络(Boundary Conditional Generative Adversarial Networks,BCGAN)的不平衡大数据模糊分类算法,通过在多数类数据和少数类数据的决策边界附近引入一个边界少数类到过样本,生成更合适的少数类数据来提高分类性能.将处理过的平衡数据转换成概率索引表,数据和属性分别以行和列的形式呈现,计算每个数据属性中存在的唯一符号的隶属度,然后设计相关模糊朴素贝叶斯(Correlative Fuzzy Naive Bayes,CFNB)分类器进行数据分类.本文给出MapReduce框架下大数据模糊分类的并行实现.实验结果表明:所提基于BCGAN的不平衡大数据模糊分类准确度优于其他现有方法,说明该方法具有可行性和有效性.

Aiming at the imbalance problem in big data classification,an unbalanced big data fuzzy classification algorithm based on boundary condition generative adversarial networks(BCGAN)has been proposed.In this method,BCGAN oversampling method is proposed by introducing a boundary minority class to oversampling near the decision boundary of majority class data and minority class data,generating more appropriate minority class data to improve the classification performance.The processed balance data is transformed into probability index table,and the data and attributes are presented in the form of row and column respectively.The membership degree of the unique symbol in each data attribute is calculated,and then the data category is obtained by means of the correlative fuzzy naive Bayes(CFNB)classifier.Then,the parallel implementation of big data fuzzy classification in MapReduce framework is given.The experimental results show that the accuracy of the proposed method is better than that of other existing methods,indicating the feasibility and effectiveness of the proposed method.