不完备决策系统基于类间容差类的属性约简算法

打开文本图片集

中图分类号：TP391 文献标志码：A 文章编号：1001-3695（2026）01-021-0178-05

doi：10.19734/j. issn.1001-3695.2026.04.0167

Attribute reduction algorithm based on inter-class tolerance classes in incomplete decision systems

Chen Chunyuan，Yin Feng†，Wu Liangkun （SchoolofComputer ScienceandArtificial Intelligence，SouthwestMinzu University，Chengdu 61Oo41，China）

Abstract：Sincetraditionalroughsettheoryignores thechangesintherelationshipbetweenobjectsindiferenttoleranceclassesandacross clusterswhendealing withincompletedecisionsystems withmisingvalues，thispaper proposedanatributereductionalgorithmbasedoninter-class tolerance.Firstly，itdividedthedomainaccording tothedecisionatributes，and definedthesamples withinthesamedecisionatributeasacluster.Itproposedtheconceptsofinter-classconsistencyanddiscriminationintheincompletedecisionsystem，andproposedtheircalculationmethodsbasedoninter-class toleranceclases. Secondly，basedontheprincipleofkeepingtheglobal inter-classdiseriminationunchanged，itproposedanewstandardfor evaluatingatributeimportance，andcalculatedthereductionincombinationwithaheuristicsearch strategy.Finall，basedon nineUCIpublicdatasets，itcomparedtheproposedalgorithm withfourother atributereductionalgorithms.Theresults show thatthe proposed algorithmcan efectivelyremoveredundantatributes，andtheaverageclassificationaccuracyisimprovedby （204 3.51% compared withthecomparisonalgorithm.Therefore，whileensuring thereductionefect，thealgorithmcaneffectively improve the classification performance of the reduction result.

Key Words：attribute reduction； incomplete decision system；inter-class tolerance classes；rough set

0 引言

粗糙集理论作为处理不确定性与信息不一致性的数学工具，现已在模式识别[1，2]决策分析[3]等领域得到广泛应用。（剩余12959字）