一种适用于多类不平衡数据集的模糊关联分类方法

A fuzzy associative classification method for multi-class imbalanced datasets

提出一种适用于多类不平衡分布情形下的模糊关联分类方法,该方法以最小化AdaBoost.M1W集成学习迭代过程中训练样本的加权分类错误率和子分类器中模糊关联分类规则数目及规则中所含模糊项的数目为遗传优化目标,实现了AdaBoost.M1W和模糊关联分类建模过程的较好融合.通过5个多类不平衡UCI标准数据集和现有的针对不平衡分类问题的数据预处理方法实验对比结果,表明了所提出的方法能显著提高多类不平衡情形下的模糊关联分类模型的分类性能.

A fuzzy associative classification method for multi-class imbalanced datasets is presented.The method implements a better combination of AdaBoost.M1W and the process of building fuzzy associative classification by the genetic optimization objective,which is minimization weighted error rate in the process of ensemble iterative learning and the number of fuzzy association rule and total fuzzy items in the weak fuzzy associative classifier.The experiments of comparing with existing data preprocessing approaches aiming at the imbalanced classification problem show that the proposed method can dramatically improve the classification performance of the fuzzy associative classifier for multi-class imbalanced datasets by five UCI multi-class imbalanced benchmark datasets.