集群分类映射的文本多标签模糊关联降维聚类  

Clustering-classification-mapping based text multi label fuzzy association dimension reduction clustering

作　　者：刘娜[1] 毛晓菊[1] 吴敏[2] 

机构地区：[1]商丘学院计算机工程学院 [2]中国科学技术大学软件学院

出　　处：《计算机工程与设计》2017年第6期1657-1663,共7页Computer Engineering and Design

基　　金：国家自然科学基金项目(61272131;61379040)

摘　　要：为实现文本的多标签分类,同时降低计算复杂度并保持分类精度,提出基于集群分类映射的文本多标签模糊关联降维聚类方法。利用模糊变换、模糊关联聚类、集群分类映射、阈值查找和应用等技术,构建低维特征的多标签模糊关联分类器的训练和测试阶段,采用模糊相关评价将高维文本转化为低维的模糊关联向量,避免维数灾难问题。所提算法不要求分类区域呈现凸性特征,适用性更广,对其进行了计算复杂度理论分析。在标准测试集上进行对比测试,测试结果验证了该算法在计算复杂度和分类精度上的优势。For multi label classification of the text,to reduce the computational complexity and maintain the classification accuracy,a text classification was put forward based on multi label fuzzy association dimension reduction clustering method of fuzzy association label.With the fuzzy transformation,fuzzy clustering,cluster classification mapping,threshold search and application,the training phase and testing phase with low dimensional feature for multi label fuzzy associative classifier were constructed,and the fuzzy relational similarity between vectors and text labels was established.The proposed algorithm does not require the classification zone to show convexity characteristic,so it can be widely applicable.The calculation complexity analysis was carried out for the proposed text classification algorithm.Through the comparison in the standard test set,the proposed algorithms show advantages in computational complexity and classification accuracy.

关 键 词：模糊变换 模糊关联聚类 集群分类映射 阈值 文本分类 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]