不均衡数据分类下特征有效识别分析

Feature Effective Recognition Analysis under Unbalanced Data Classification

不均衡数据分类下对其特征进行高精度识别,能够有效提高数据的均衡分类精度,改善数据混乱现象。对数据分类过程下的特征识别,需要先获得非均衡数据功率谱密度函数特征,对传输的非均衡数据进行特征筛选,完成数据特征的有效识别。传统方法利用函数极值法确定隶属度,获取不均衡数据分类过程的特征参数,但忽略了对数据特征进行进一步择优筛选,导致了数据特征识别精度偏低。提出不均衡数据分类下的特征有效识别方法。先构建了数据传输过程中非均衡数据信息特征模型,获得传输非均衡数据功率谱密度函数特征,并对传输的非均衡数据进行特征筛选。采用选择性集成方式对传输过程中的非均衡数据基分类器进行学习,解决各个分类器集合之间的模糊差异度问题,完成不均衡数据分类中特征高效识别。实验结果表明,所提方法有效提高非均衡数据分类效率,加快了不均衡数据处理速度。

The aim of this article is overcome defect of traditional feature recognition method under unbalanced data classification, such as poor recognition precision. Under unbalanced data classification, a new recognition meth- od is presented. Firstly, model of information feature of unbalanced data is built during data transmission to obtain feature of power spectral density function transferring unbalanced data and feature screening is carried out for the un- balanced data. Then, selective integrated approach is used to study the classifier based on unbalanced data and de- gree of fuzzy difference between each classifier set is solved. Thus, the efficient recognition of feature is completed. Experimental results show that the proposed method can improve classification effectiveness effectively. It accelerates processing speed of the unbalanced data.