邻域决策的随机约简与集成分类研究  

Neighborhood Based Randomized Reduction and Ensemble Classification Approach

作　　者：余思成 杨习贝[1,2] 陈向坚[1] 窦慧莉[1] 王平心[3] YU Si-cheng;YANG Xi-bei;CHEN Xiang-jian;DOU Hui-li;WANG Ping-xin(School of Computer Science, Jiangsu University of Science and Technology, Zhenjiang 212003, China;School of Economics ＆ ent, Nanjing University of Science and Technology, Nanjing 210094, China;School of Mathematics and Physics, Jiangsu University of Science and Technology, Zhenjiang 212003, China)

机构地区：[1]江苏科技大学计算机学院,江苏镇江212003 [2]南京理工大学经济管理学院,南京210094 [3]江苏科技大学数理学院,江苏镇江212003

出　　处：《小型微型计算机系统》2018年第6期1163-1167,共5页Journal of Chinese Computer Systems

基　　金：国家自然科学基金项目(61572242;61503160;61502211)资助;江苏省高校哲学社会科学基金项目(2015SJD769)资助;中国博士后科学基金项目(2014M550293)资助

摘　　要：为了进一步提升邻域分类器的性能,提出基于属性约简的集成邻域分类策略.首先在启发式求解约简的过程中,通过放宽属性选择的条件,从而在一定范围内利用随机选择的方法获取多个能够降低邻域决策错误率的属性子集,然后借助这些属性子集在对应邻域分类器上得到的结果进行投票集成,得到最终的分类类别.在12个UCI数据集上的实验结果表明,所提出的基于属性约简的集成邻域分类策略不仅能够有效地提升邻域分类器的分类精度,而且亦能增强邻域分类结果的鲁棒性.这一研究为从集成的视角研究粗糙集理论提供了技术支持.To further improve the performance of neighborhood classifier,an ensemble strategy for neighborhood classifier is proposed based on randomized reduction. Firstly,a random parameter is introduced into the heuristic process for computing multi-different reducts that reduce the decision error of neighborhood classifier. Secondly,the voting ensemble is employed for fusing the neighborhood classification results by these reducts. The experimental results on 12 UCI data sets tell us the proposed strategy can improve not only the classification accuracy but also the classification robustness. This study provides us a technique for studying rough set theory via ensemble learning strategy.

关 键 词：集成学习 一致性 邻域分类器 邻域决策错误率 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]