感性工程Kansei EngineeringDay4:KanseiRecommendationNovember23,2017YoshiteruNAKAMORl,ProfessorEmeritusJapanAdvanced Institute of ScienceandTechnology
Kansei Engineering Day 4: Kansei Recommendation November 23, 2017 Yoshiteru NAKAMORI, Professor Emeritus Japan Advanced Institute of Science and Technology 感性工程
Today'sContents1.OrderedWeightedAveragingMethod有序加权平均2.Exercise3:ProductRecommendation产品推荐3.Non-additiveAggregationMethod非加性聚集4.Exercise4:ProductRecommendation产品推荐CheerfulMelancholy快乐愁绪<
1. Ordered Weighted Averaging Method 有序加权平均 2. Exercise 3: Product Recommendation 产品推荐 3. Non-additive Aggregation Method 非加性聚集 4. Exercise 4: Product Recommendation 产品推荐 Today’s Contents Cheerful 快乐 2 Melancholy 愁绪
1. Ordered Weighted Averaging MethodPleaseshowmesomeceramiccoffeecups,whicharesomewhatcute,verymodernbutnotveryexpensive.OI want a cupthat meets mydesiresasmanyaspossibleW:SomewhatcuteW2:Verymodern.MeetsallofthedesiresW3:Notveryexpensive·MeetsthedesiresasmanyaspossibleMeetsalleasthalfofthedesiresCupstoberecommendedaredifferent.Meets atleast onedesire
1. Ordered Weighted Averaging Method Please show me some ceramic coffee cups, which are somewhat cute, very modern, but not very expensive. I want a cup that meets my desires as many as possible. Cups to be recommended are different. 3 • Meets all of the desires. • Meets the desires as many as possible. • Meets al least half of the desires. • Meets at least one desire. w1 w2 w3 ： Not very expensive ： Somewhat cute ： Very modern
(Example) Which do you recommend?0102S22=0.9 Very modernEvaluationvalueNot very expensive Sis=0.7S2, =0.6 Somewhat cuteSomewhat cute Su=0.5Very modern Sj2 =0.3Totalscore=1.5Total score=1.6一S23 = 0.1 Not very expensive0102·Meetsall ofthedesires.0102>.Meetsthedesiresasmanyas possible.>0102.Meets al leasthalf ofthedesires.>0102.Meetsatleastonedesire
o2 o2 o2 o2 o1 o1 o1 o1 (Example) Which do you recommend? 4 Somewhat cute Very modern Not very expensive Somewhat cute Very modern Not very expensive Evaluation value • Meets all of the desires. • Meets the desires as many as possible. • Meets al least half of the desires. • Meets at least one desire. Total score =1.5 Total score =1.6 o1 o2 s11 = 0.5 0.3 s12 = 0.7 s13 = s21 = 0.6 0.9 s22 = s23 = 0.1
wy:Somewhatcuteldea of ordered weighted averagingWz:Verymodernw:Notveryexpensive·Sortindescendingorderofattributeevaluationvalues:Sn =0.5 S12 =0.3 S3 = 0.7S = 0.7 si2 = 0.5 si, = 0.3.Overall evaluation by weighted average重要！j,2,HowtodetermineweightsE=x+μs2+s3.Whentheaverage value (orthetotal)istakenastheoverall evaluation1Inthefieldofdecision(or 1.0)===analysis,"Somewhatcute"3is calledanattribute,and·Exampleonthepreviouspage:O,hasalargeraverage.the scoreiscalled anattributeevaluationvalue
Idea of ordered weighted averaging • Sort in descending order of attribute evaluation values: • Overall evaluation by weighted average • When the average value (or the total) is taken as the overall evaluation • Example on the previous page: has a larger average. s11 = 0.5 s12 = 0.3 s13 = 0.7 s11 ′ = 0.7 s12 ′ = 0.5 s13 ′ = 0.3 1 1 11 2 12 3 13 E = µ × s ′ + µ × s ′ + µ × s ′ (or 1.0) 3 1 µ1 = µ2 = µ3 = 5 1 2 3 µ ,µ ,µ How to determine weights In the field of decision analysis, “Somewhat cute" is called an attribute, and the score is called an attribute evaluation value. o2 w1 w2 w3 ： Not very expensive ： Somewhat cute ： Very modern
o02Weight (1)=0.9VerymodemNot very expensive Su, =0.7=0.6SomewhatcuteSomewhat cute S,,=0.5Very modern Sμz = 0.3.MeetsallofthedesiresTotalscore =1.5Total score =1.6S, = 0.1 Not very expensive.The lowest score among the attribute evaluation values is the evaluation value of the object(s". ≥s'm, ≥s"m.Em=μ×sm+μ,×sm2+u,×sm3H2Consider onlythe smallestμ =0.0, μ2 = 0.0, μ =1.0attribute value..Then,theobject withthehighestscoreamongtheaboveevaluationscores is selected.Intheexampleabove:O,isselectedbecauseE, =0.3, E, =0.1Max-minstrategy:Astrategyusuallyusedinriskmanagement,asolidstrategy
Weight (1) • Meets all of the desires • The lowest score among the attribute evaluation values is the evaluation value of the object • Then, the object with the highest score among the above evaluation scores is selected. In the example above: is selected because • Max-min strategy : A strategy usually used in risk management, a solid strategy µ1 = 0.0, µ2 = 0.0, µ3 =1.0 ( ) 1 1 2 2 3 3 1 2 3 m m m m m m m E = µ × s ′ + µ × s ′ + µ × s ′ s ′ ≥ s ′ ≥ s ′ E1 = 0.3, E2 = 0.1 o1 6 Consider only the smallest attribute value
o0.Weight (2),=0.9VerymodemNot very expensive Su, =0.7=0.6SomewhatcuteSomewhat cute S,,=0.5Very modern Sμz = 0.3.MeetsatleastonedesireTotalscore =1.5Total score =1.6S,, = 0.1 Not very expensive.The highest score among the attribute evaluation values is the overall evaluation value(s". ≥s', ≥s".Em=μ×sm+μ,×sm2+u,×sm32Consideronlythelargestμ =1.0, μz = 0.0, μ, = 0.0attribute value..Then,the object with thehighestscoreamong theabove evaluation scores is selected.Intheexampleabove:O,isselectedbecauseE, =0.7, E, =0.9Max-maxstrategy:Riskisbig,butifyouhititwell yougetabigreturn
Weight (2) • Meets at least one desire • The highest score among the attribute evaluation values is the overall evaluation value • Then, the object with the highest score among the above evaluation scores is selected. In the example above: is selected because • Max-max strategy: Risk is big, but if you hit it well you get a big return. µ1 =1.0, µ2 = 0.0, µ3 = 0.0 ( ) 1 1 2 2 3 3 1 2 3 m m m m m m m E = µ × s ′ + µ × s ′ + µ × s ′ s ′ ≥ s ′ ≥ s ′ E1 = 0.7, E2 = 0.9 o2 7 Consider only the largest attribute value
Weight (3)0102S22=0.9VerymodernEvaluationvalueNotveryexpensiveSi,=0.7S2,=0.6 Somewhat cuteSomewhat cuteS.=0.5Verymodern S,=0.3Total score=1.5Total score =1.6S23 =0.1 Not very expensive21.11E.x0.5x0.3.Meets the desires as many as possible.0133320.8E,×0.6+x0.1:μj= 0.0, μ2 =1/3, μ3 =2 /3333211.9.Meetsal leasthalfof thedesires.02E,x0.5x0.7333212.4μ= 2 /3, μ2 =1/3, 3 = 0.0×0.6=E, =0.9+333
Weight (3) • Meets the desires as many as possible. • Meets al least half of the desires. µ1 = 0.0, µ 2 = 1/ 3, µ3 = 2 / 3 2 / 3, 1/ 3, 0.0 µ1 = µ 2 = µ3 = 1 o o2 8 3 0.8 0.1 3 2 0.6 3 1 3 1.1 0.3 3 2 0.5 3 1 2 1 = × + × = = × + × = E E Somewhat cute Very modern Not very expensive Somewhat cute Very modern Not very expensive Evaluation value Total score =1.5 Total score =1.6 o1 o2 0.5 s11 = s12 = 0.3 s13 = 0.7 s21 = 0.6 s22 = 0.9 0.1 s23 = 3 2.4 0.6 3 1 0.9 3 2 3 1.9 0.5 3 1 0.7 3 2 2 1 = × + × = = × + × = E E
Another examplePleaseshowmesomeceramiccoffeecups, which are somewhat cute,verymodern,butnotveryexpensiveCupO3Cup OsCup 02Cup 04Cup O1
Another example 9 Cup o2 Cup Cup o4 o3 Cup Cup o1 o5 Please show me some ceramic coffee cups, which are somewhat cute, very modern, but not very expensive
Evaluation of coffee cups·HerdesiresareSomewhatcute,Verymodern'andNotveryexpensive'(say,3,oooYen).·SheevaluatedthesefivecupsusingherKansei,andtheopinionof theshopowner.TastefulCuteTraditional←→ModernExpensive←→CheapWiW2 (Opinion of the owner)W3 (Objective value)(Her Kansei)Cup (o))SomewhattastefulorNeutralModern￥6,000Cup (02)Cute￥5,500NeutralorSomewhatmodernCup (03)￥1,500TastefulorSomewhattastefulSomewhatmodernCup (04)VerycuteSomewhattraditionalorNeutral￥3,500Cup (os)￥6,000TastefulorSomewhattastefulModernorVerymodern10
Evaluation of coffee cups • Her desires are ‘Somewhat cute’, ‘Very modern’ and ‘Not very expensive’ (say, 3,000 Yen). • She evaluated these five cups using her Kansei, and the opinion of the shop owner. Tasteful Cute (Her Kansei) Traditional Modern (Opinion of the owner) Expensive Cheap (Objective value) Cup ( ) Somewhat tasteful or Neutral Modern ¥6,000 Cup ( ) Cute Neutral or Somewhat modern ¥5,500 Cup ( ) Tasteful or Somewhat tasteful Somewhat modern ¥1,500 Cup ( ) Very cute Somewhat traditional or Neutral ¥3,500 Cup ( ) Tasteful or Somewhat tasteful Modern or Very modern ¥6,000 10 o1 o2 3 o o4 5 o w1 w2 w3