基于柔性逻辑的属性约简算法的改进及其对比  

Improvement and comparison of attribute reduction algorithms based on flexible logic

作　　者：刘城霞[1] 朱敏玲[2] LIU Chengxia ZHU Minling(Beijing Key Laboratory of Internet Culture ＆ Digital Dissemination Research, Beijing Information Science and Technology University, Beijing 100101, China Computer School, Beijing Information Science ＆ Technology University, Beijing 100101, China)

机构地区：[1]北京信息科技大学网络文化与数字传播北京市重点实验室,北京100101 [2]北京信息科技大学计算机学院,北京100101

出　　处：《北京信息科技大学学报（自然科学版）》2017年第5期17-23,共7页Journal of Beijing Information Science and Technology University

基　　金：2017年网络文化与数字传播北京市重点实验室开放课题资助;2017北京高等学校高水平人才交叉培养"实培计划"毕业设计(论文)项目资助

摘　　要：在数据挖掘中的连续数据往往是需要先离散化处理后才能进行挖掘,而离散化方法的好坏也影响着最后的挖掘效果。柔性逻辑是一种可以包容不同逻辑的逻辑谱系,它可以直接对连续数据进行处理。如此将柔性逻辑中的柔性等价作为新的不可分辨关系,并利用等价类的概念和离散定律中吸收率的概念对不可分辨函数进行化简,进而进行改进后的属性约简,将是一种新的更直观、更符合实际数据要求的约简方法。基于这种思路改进了属性约简的算法,并用实验数据比较了改进前后的约简时间的差异。从结果来看应用新不可分辨关系在进行数据处理时省去了数据离散化过程,简化了数据预处理的步骤,有效地减少属性约简的时间。另外,为了和已有属性约简算法进行横向比较验证,还用实验数据验证了基于可分辨矩阵属性约简、启发式的属性约简、基于条件信息熵的属性约简及动态属性约简等经典属性约简算法,并将其和改进属性约简算法的约简时间进行了对比。The continuous attributes need to be discretized in data mining and the discretization method affects the mining efficiency. The flexible logic is a logical spectrum which can include different logics and it can process the continuous data directly. By combining the flexible equation with the indiscernible relation,the new improved indiscernible relation is got. In addition,the concept of equivalence class and absorption of discrete mathematics are used to simplify the indiscernible function for its attribute reduction. The reduction operating time is compared by the experiment data. The result shows obvious operating time reduction because application of new indiscernible relation simplifies the data pre-processing by omission of data discretion. To compare with the existing attribute reduction algorithms horizontally,the paper realizes the attribute reduction based on discernibility matrix,heuristic attribute reduction, the attribute reduction based on information entropy, and dynamic attribute reduction. The operating time using these classical algorithms is compared with the new method improved by indiscernibility correlation and based on discernibility matrix.

关 键 词：数据挖掘 属性约简 柔性逻辑 不可分辨关系 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]