论文部分内容阅读
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.
This paper proposes a fuzzy relevance classification method which is suitable for many kinds of unbalanced distribution cases. This method minimizes the weighted classification error rate of training samples and the number and rules of fuzzy relevance classification rules in AdaBoost.M1W integrated learning iterative process The number of fuzzy items included is the goal of genetic optimization, which achieves a better fusion of AdaBoost.M1W and fuzzy association classification modeling process.Through the five UCI standard datasets and the existing data for unbalanced classification Experimental results of the pretreatment method show that the proposed method can significantly improve the classification performance of the fuzzy correlation classification model under multiple types of imbalances.