資源描述:
《基于聚類的多屬性群決策專家權(quán)重確定方法》由會員上傳分享,免費(fèi)在線閱讀,更多相關(guān)內(nèi)容在學(xué)術(shù)論文-天天文庫。
1、第23卷第6期運(yùn)籌與管理Vol.23,No.62014年12月OPERATIONSRESEARCHANDMANAGEMENTSCIENCEDec.2014基于聚類的多屬性群決策專家權(quán)重確定方法112何立華,王櫟綺,張連營(1.中國石油大學(xué)(華東)經(jīng)濟(jì)管理學(xué)院,山東青島266580;2.天津大學(xué)管理與經(jīng)濟(jì)學(xué)部,天津300072)摘要:對于多屬性群決策中專家權(quán)重確定的問題,本文提出了基于聚類的專家權(quán)重確定方法,將專家權(quán)重分為類別間權(quán)重和類別內(nèi)權(quán)重,對專家聚類步驟和類別間權(quán)重的計算方法進(jìn)行了改進(jìn)。通過專家給出的判斷矩陣構(gòu)建相容度矩陣,利用系統(tǒng)聚類
2、原理,對相容度矩陣進(jìn)行聚類,得到最大相容度譜系圖。通過最大相容度間的距離和給定閾值的比較,對專家進(jìn)行恰當(dāng)分類,從而避免了根據(jù)現(xiàn)有研究步驟只能將專家分為兩類的不足。此外,在確定類別間權(quán)重時,除繼續(xù)對類容量較大的類賦予較大的類別間權(quán)重系數(shù)外,還引入專家判斷矩陣的屬性權(quán)重一致性來反映類別間的差異,從而有效避免了當(dāng)某幾類專家中含有相等數(shù)目專家時,賦予這幾類專家相同類別間權(quán)重系數(shù)的問題。所提方法結(jié)構(gòu)清晰、計算簡便,并使得專家權(quán)重計算結(jié)果更為合理準(zhǔn)確。最后運(yùn)用一個算例對比驗證了該方法的可行性和有效性。關(guān)鍵詞:決策科學(xué);多屬性群決策;專家權(quán)重;聚類分析;
3、判斷矩陣中圖分類號:C934文章標(biāo)識碼:A文章編號:1007-3221(2014)06-0065-08AMethodforDeterminingtheExperts’WeightsofMulti-AttributeGroupDecision-MakingBasedonClusteringAnalysis112HELi-hua,WANGLi-qi,ZHANGlian-ying(1.SchoolofEconomicsandManagement,ChinaUniversityofPetroleum(EastChina),Qingdao266580
4、,China;2.CollegeofManagementandEconomics,TianjinUniversity,Tianjin300072,China)Abstract:Anexperts’weightdeterminingmethodbasedontheexperts’weightsclusteringanalysisisproposedtodeterminetheexperts’weightsofmulti-attributegroupdecision-making.Theexperts’weightisdividedintoth
5、eweightsbetweencategoriesandwithincategory.Thestepsofexperts’clusteringandthecalculationmethodoftheweightsbetweencategoriesareimproved.Theclusteringpedigreechartofthemaximumcompatibilityde-greeisgotbybuildingtheexpertjudgmentcompatibilitymatrixaccordingtotheexpertjudgmentm
6、atrix,makinguseofthesystemclusteringprincipletoclusterthecompatibilitydegreematrix.Theexpertsareclassifiedproper-lyaccordingtothecomparisonofthedistancebetweenthemaximumcompatibilitydegreeandthegiventhresholdvalue,whichovercomestheshortcomingofonlyclusteringtheexpertsintot
7、wocategoriesinexistingliteratures.Inaddition,whiledeterminingtheweightsbetweencategories,itnotonlygivesgreaterweightforthelargerca-pacity,butalsoreflectsthedifferencebetweencategoriesbyintroducingtheattribute’sweightconsistenceoftheexperts’judgmentmatrix,whichavoidsgivingt
8、hesameweightsbetweencategoriesforthosecategoriesthathavethesamenumberexperts.Thestructure