基于分类规则挖掘的数据多标记特征分层识别  

Hierarchical Recognition of Data Multi Label Features Based on Classification Rule Mining

作　　者：朱方娥 郭建方 曹丽娜 ZHU Fang-e;GUO Jian-fang;CAO Li-na(Sifang College,Shijiazhuang Tiedao University,Shijiazhuang Hebei 051132,China)

机构地区：[1]石家庄铁道大学四方学院,河北石家庄051132

出　　处：《计算机仿真》2021年第4期310-314,共5页Computer Simulation

摘　　要：面对多标签的大数据集,传统的分类识别方法识别质量不高,现提出基于分类规则挖掘的数据多标记特征分层识别方法。利用改进后的LLE(局部线性嵌入)方法进行数据降维处理,搜索数据特征,并对数据多标记特征选择,构成特征子集,根据特征子集,利用分类规则挖掘方法构建一个分类识别模型,实现数据多标记特征分层识别。结果表明,与传统方法相比,所研究方法识别下,汉明损失度最小,数据多标记特征分层识别准确度最大,说明上述方法的识别质量较高,达到了研究的预期目标,为数据利用和挖掘提供了参考和借鉴。The current classification methods have low recognition quality when dealing with large multi label data sets.Therefore,this paper proposes a new hierarchical recognition method based on classification rule mining.The improved local linear embedding(LLE)method was applied to reduce dimension data,search data features and select multi-label features to form feature subset.Based on feature subset,a classification recognition model was constructed by using classification rule mining method thus realizing multi-label feature hierarchical recognition.The results show that the method has low Hamming loss,high accuracy and excellent recognition quality,providing reference for data utilization and mining.

关 键 词：分类规则挖掘 多标签数据 数据降维处理 多标记特征选择 分层识别方法 免疫算法 

分 类 号：TP184.66[自动化与计算机技术—控制理论与控制工程]