南京大学学报（自然科学版）Issue(4)：880-893,14.DOI:10.13232/j.cnki.jnju.2015.04.026
基于 Rough Set 的高维特征选择混合遗传算法研究
A new hybrid genetic algorithm for high dimension feature selection based on rough set
摘要
Abstract
Genetic algorithm is an effective method to resolve attribute reduct problem,which is a NP-hard problem. Fitness function is one of the key issues.Focusing on this problem,a new Hybrid Genetic Algorithm(HGA)for high dimension feature selection based on rough set(RS)is putted forward.From the following two aspects,algebra and information entropy of rough set,HGA comprehensively takes into account some factors of attribute reduction set, such as the number of attributes,chromosome coding,genes value,attribute importance,attribute dependency, attribute relevancy.First,a hybrid common fitness function framework of genetic algorithm is proposed.And then, different fitness function are realized by adjusting weight coefficients of various factors.Finally,one hundred and two features of MRI prostate tumor ROI are extracted to build decision information table of prostate tumor patients.Four experiments of high-dimensional features selection are carried out,using neural network to verify influence degree of recognition accurancy in different fitness function parameters.Experimental results show that the algorithm is effective.However,there are greater impacts using different parameters,and appropriate parameters should be used to different problem in order to obtain better recognition accurance.关键词
粗糙集/特征约简/遗传算法/属性依赖度/属性重要度Key words
rough set/attribute reduct/genetic algorithm/attribute dependency/attribute importance分类
信息技术与安全科学引用本文复制引用
周涛,陆惠玲,张艳宁,马苗..基于 Rough Set 的高维特征选择混合遗传算法研究[J].南京大学学报（自然科学版）,2015,(4):880-893,14.基金项目
国家自然科学基金(81160183),宁夏省自然科学基金(NZ12179,NZ14085),宁夏高等学校科研项目(NGY2013062),陕西省语音与图像信息处理重点实验室开放课题(SJ2013003),宁夏医科大学特殊人才项目(XT2011004) （81160183）
