基于时序数列的动态多属性决策方法分析及应用-analysis and application of dynamic multi-attribute decision-making method based on time series.docxVIP

摘要决策是从所有选择的方案中找出决策者最满意的一种方法，作为实现预定目标的策略，应用广泛。实际问题研究系统的复杂，需要从多个角度（属性、目标、指标、时间）进行分析，导致多属性决策在决策过程中的不足。本文根据现实中存在的问题，引入时间因素偏好概念，提出动态多属性决策问题，系统地将历史数据、现实状态和将来预测的结果综合集成，并且对决策结果提出证据理论的决策研究。本文研究的主要创新成果如下：（1）在多属性决策问题中，引入时间偏好序列概念，构建一个基于OWA的动态多属性决策模型，从属性、指标和时间三维空间的角度综合考虑决策结果；（2）将原来对历史数据、现在时间点的决策延伸至对历史数据、现在时间点和将来数据的综合考虑，具有一定的创新性，对很多现实中存在的复杂问题的决策具有重要的理论指导意义；（3）对将来数据的引入，构建了一个基于精度优化的预测模型，在整体决策中将预测结果的精度有效地映射到最终综合评价结果中；（4）整体将历史数据、现在节点和将来数据集成，构建证据推理的精度优化模型，将决策结果给出信任区间，得到最可能的方案。对上述几类问题的研究，均给出了算例，验证了模型的有效性，为今后进一步研究打下基础。关键词：多属性决策，时序数列，OWA,动态多属性决策，证据理论ABSTRACTDecision-makingasastrategytoachievetheintendedgoals,fromalltheiroptionstofindthemostsatisfactorymethodofdecision-makers,isappliedofthewiderange.Thepracticalproblemsofthecomplexityofthesystemneedtobeanalyzedfromvariousangles(attributes,criterias,targets,time),resultinginthelackofMultipleCriteriaDecisioninthedecision-makingprocess.Accordingtothereality,thisarticleintroducesthetimefactorpreference,putsforwardMultiple-Criteriaproblems,systematicallyconsideringhistoricaldata,realstateandthecomprehensivestudyofthefuturepredictingresults,anddiscussestheevidencetothetheoryofdecision-makingresults.Themaininnovativepointsareasfollows:IntheMultipleCriteriaDecisionAnalysisproblems,introducetheconceptoftime,tobuildaOWA-baseddynamicMultipleCriteriaDecisionAnalysismodel,consideringthedecisionresultsfromthepointofviewofthethree-dimensionalspaceoftheattributes,indicatorsandtime;Extendsthedecision-makingoforiginalhistoricaldataandthetimepointtothecomprehensiveconsideringofthehistoricaldata,thepoint-in-timeandfuturedata,whichhasacertainsenseofinnovation,andhasimportanttheoreticalguidingsignificanceinthedecision-makingofmanycomplexissuesofreality;Introducethefuturedatatoconstructapredictivemodelbasedonaccuracyoptimization,andtheaccuracyofthepredictedresultsintheoveralldecision-makingwillbemappedtothefinalresultsofthecomprehensiveevaluation;Integratehistoricaldata,nownodesandfuturedatatobuildtheaccuracyoptimizationmodeloftheevidentialreasoning,giventhecon