1 / 86

Roman Słowiński Poznań University of Technology, Poland

Multiple-criteria ranking using an additive value function constructed via ordinal regresion : UTA method. Roman Słowiński Poznań University of Technology, Poland.  Roman Słowiński. g 2 ( x ). g 2max. A. g 2min. g 1 ( x ). g 1min. g 1max. Problem statement.

alka
Télécharger la présentation

Roman Słowiński Poznań University of Technology, Poland

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Multiple-criteria ranking using an additive value function constructed via ordinal regresion :UTA method Roman Słowiński Poznań University of Technology, Poland  Roman Słowiński

  2. g2(x) g2max A g2min g1(x) g1min g1max Problem statement • Consider a finite set A of actions (actions, solutions, objects) evaluated by m criteria from a consistent family F={g1,...,gm} • Let I={1,…,m}

  3. What is a consistent family of criteria ? • A family of criteria F={g1,...,gn} is consistent if it is: • Complete – if two actions have the same evaluations on all criteria, then they have to be indifferent, i.e. if for any a,bA, there is gi(a)~gi(b), iI, then a~b • Monotonic – if action a is preferred to action b (ab), and there is action c, such that gi(c)gi(a), iI, then cb • Non-redundant – elimination of any criterion from the family F should violate at least one of the above properties

  4. A x * * x * * * x x * * x x x * * * x x x x x x * * x x x x Problem statement • Taking into account preferences of a Decision Maker (DM), rank all the actions of set A from the best to the worst x

  5. g2(x) g2max nadir A ideal g2min g1(x) g1min g1max Dominance relation • Action aA is non-dominated (Pareto-optimal) if and only if there is no other action bA such that gi(b)gi(a), iI, and on at least one criterion jI, gi(b)gi(a)

  6. Criteria aggregation model = preference model • Dominance relation is too poor – it leaves many actions non-comparable • One can „enrich” the dominance relation, using preference information elicited from the Decision Maker • Preference information permits to built a preference model that aggregates the vector evaluations of elements of A

  7. Why traditional MCDM methods may confuse their users ? • Traditional MCDM methods require a rich and difficult preference information: • many intracriteria and intercriteria parameters: thresholds, weights, … • complete set of pairwise comparisons of actions on each criterion • complete set of pairwise comparisons of criteria • … • They suppose the DM understands the logic of a particular aggregation model: • meaning of weights: substitution ratios or relative strengths • meaning of lotteries (ASSESS) • meaning of indifference, preference and veto thresholds (ELECTRE) • meaning of the ratio scale of the intensity of preference (AHP) • meaning of „neutral” and „good” levels on particular criteria (MACBETH) • …

  8. Towards „easy” preference information • Traditional methods appear to be too demanding of cognitive effort of their users • This is why we advocate for methods requiring „easy” preference information • „Easy” means natural and even partial • Psychologists confirm that DMs are more confident exercising their decisions than explaining them

  9. A Towards „easy” preference information • The most natural is a holistic pairwise comparison of some actions relatively well known to the DM, i.e. reference actions

  10. holistic preference information A xy zw xw yv ut zu uz DM y x AR t z v w u Towards „easy” preference information • The most natural is a holistic pairwise comparison of some actions relatively well known to the DM, i.e. reference actions

  11. preference information A xy zw xw yv ut zu uz Preference model compatible with preference information DM analyst y x AR t z v w u Apply the preference model on A Towards „easy” preference information • Question: what is the consequence of using on the whole set Athis information transformed to a compatible preference model ? What ranking will result ?

  12. Aggregation paradigms • Disaggregation-aggregation (or regression) paradigm: The holistic preference on a subset ARAis known first, and then a compatible criteria aggregation model (compatible preference model)is inferred from this information to be applied on set A • Traditional aggregation paradigm: The criteria aggregation model (preference model) is first constructed and then applied on set A to get information about holistic preference

  13. Aggregation paradigms • The disaggregation-aggregation paradigam has been introduced to MCDS by Jacquet-Lagreze & Siskos(1982) in the UTA method: the inferred criteria aggregation model is the additive value function with piecewise-linear marginal value functions • The disaggregation-aggregation paradigam is consistent with the „posterior rationality” principle by March(1988) and „learning from examples” used in AI and knowledge discovery • Other aggregation models inferred in this way: • Fishburn (1967) – trade-off weights • Mousseau & Słowiński(1998) – outranking relation (ELECTRE TRI) • Greco, Matarazzo & Słowiński(1999) – decision rules or trees (DRSA – Dominance-based Rough Set Approach)

  14. Basic concepts and notation • Gi – domain of criterion gi (Gi is finite or countably infinte) • – evaluation space • x,yG – profiles of actions in evaluation space • – weak preference(outranking) relation onG:for each x,yG xy„x is at least as good as y” xy [xyand notyx] „x is preferred to y” x~y [xyand yx] „x is indifferent to y”

  15. Reminder of the UTA method (Jacquet-Lagreze & Siskos, 1982) • For simplicity: Gi, iI, where I={1,…,m} • For eachgi,Gi=[i, i] is the criterion evaluation scale, i  i , where i and i, are the worst and the best (finite) evaluations, resp. • Thus, A is a finite subset of G and • Additive value(or utility) function on G: for each xG where ui are non-decreasing marginal value functions, ui : Gi, iI

  16. x1 A  x2  AR x3x4  x5  x6x7 Principle of the ordinal regression - UTA (Jacquet-Lagreze & Siskos, 1982) • The preference information is given in form of a complete preorder on a subset of reference actionsARA, AR={x1,x2,...,xn} – the reference actions are rearranged such that xkxk+1 , k=1,...,n-1

  17. a1  a2  a3 Principle of the ordinal regression • Example: Let AR={a1, a2, a3}, G={Gain_1, Gain_2} Evaluation of reference actions on criteria Gain_1, Gain_2: Reference ranking:

  18. Principle of the ordinal regression

  19. Principle of the ordinal regression

  20. a1 a1   a3 a2   a2 a3 Principle of the ordinal regression • Let’s change the reference ranking: • One linear piece per each marginal value function u1, u2 is not enough u1=k1Gain_1, u2=k2Gain_2, U=u1+u2 For a1a3, k2>k1, but for a3a2, k1>k2, thus, marginal value functions cannot be linear

  21. Principle of the ordinal regression

  22. Principle of the UTA method (Jacquet-Lagreze & Siskos, 1982) • The comprehensive preference information is given in form of a complete preorder on a subset of reference actionsARA, AR={x1,x2,...,xn} – the reference actions are rearranged such that xkxk+1 , k=1,...,n-1 • The inferred value of each reference action xAR where + and - are potential errors of over- and under-estimation of the right value, resp. • The intervals [i, i] are divided into iequal sub-intervals with the end points (iI)

  23. Principle of the ordinal regression • In the UTA method, the marginal value of action xA is approximated by linear interpolation: for

  24. UTA additive preference model

  25. Principle of the UTA method • Ordinal regression principle for xkxk+1 , k=1,...,n-1 • Monotonicity of preferences • Normalization

  26. k=1,...,n-1 (C) Principle of the UTA method • The marginal value functions (breakpoint variables) are estimated by solving the LP problem where  is a small positive constant

  27. polyhedron of constraints (C) EUTA EUTA*+ EUTA= EUTA* Principle of the UTA method • If EUTA*=0, then the polyhedron of feasible solutions for ui(xi) is not empty and there exists at least one value functionU[g(x)] compatible with the complete preorder on AR • If EUTA*>0, then there is no value functionU[g(x)] compatible with the complete preorder on AR– three possible moves: • increasing the number of linear pieces i for ui(xi) • revision of the complete preorder on AR • post-optimal search for the best function with respect to Kendall’s  in the area EUTA EUTA*+  Jacquet-Lagreze & Siskos (1982)

  28. Współczynnik Kendalla • Do wyznaczania odległości między preporządkami stosuje się miarę Kendalla • Przyjmijmy, że mamy dwie macierze kwadratowe R i R* o rozmiarze m m, gdzie m = |AR|, czyli m jest liczbą wariantów referencyjnych • macierz R jest związana z porządkiem referencyjnym podanym przez decydenta, • macierz R* jest związana z porządkiem dokonanym przez funkcję użyteczności wyznaczoną z zadania PL (zadania regresji porządkowej) • Każdy element macierzy R, czyli rij (i, j=1,..,m), może przyjmować wartości: • To samo dotyczy elementów macierzy R* • Tak więc w każdej z tych macierzy kodujemy pozycję (w porządku) wariantu a względem wariantu b

  29. Współczynnik Kendalla • Następnie oblicza się współczynnik Kendalla: gdzie dk(R,R*) jest odległością Kendallamiędzy macierzami R i R*: • Stąd -1, 1 • Jeżeli  = -1, to oznacza to, że porządki zakodowane w macierzach R i R*są zupełnie odwrotne, np. macierz R koduje porządek a  b  c  d, a macierz R* porządek d  c  b  a • Jeżeli  = 1, to zachodzi całkowita zgodność porządków z obydwu macierzy. W tej sytuacji błąd estymacji funkcji użyteczności F*=0 • W praktyce funkcję użyteczności akceptuje się, gdy  0.75

  30. 1 2 Example of UTA+ • Ranking of 6 means of transportation

  31. Preference attitude: „economical”

  32. Preference attitude: „hurry”

  33. Preference attitude: „hurry”

  34. preference information xy zw yv ut zu uz All instances ofpreference model compatible with preference information A DM analyst y x AR t z v w u Apply all compatible instances on A One should use all compatible preference models on set A • Question: what is the consequence of using all compatible preference models on set A ? What rankings will result ?

  35. z u Includesnecessary ranking and does not include the complement of necessary ranking x w xy zw yv ut zu uz preference information y t v possible ranking necessary ranking Two rankings result: necessary and possible

  36. z u Includes necessary ranking and does not include the complement of necessary ranking w x xy zw yv ut zu uz xw additional preference information y t v necessary rankingenriched possible rankingimpoverished Two rankings result: necessary and possible – effect of additional preference information „trial-and-error” interactions • In the absence of any preference information: • necessary ranking boils down to weak dominance relation • possible ranking is a complete relation • For complete pairwise comparisons (complete preorder in A): • necessary ranking = possible ranking

  37. The UTAGMS method(Greco, Mousseau & Słowiński 2004) • The preference information is a partial preorder on a subset of reference actions ARA • A value function is called compatible if it is able to restore the partial preorder reference actions from AR • Each compatible value function induces a ranking on set A • In result, one obtains two rankings on set A, such that for any pair of actions (x,y)A: • x N y:x is ranked at least as good as y iff U(x)U(y) for all value functions compatible withthe preference information (necessary weak preference relation N - a partial preorder on A) • x P y:x is ranked at least as good as y iff U(x)U(y) for at least one value function compatible with the preference information (possibleweak preference relation P - a strongly complete and negatively transitive binary relation on A)

  38. ui(xi) gi 0 i i yi vi wi zi The UTAGMS method(Greco, Mousseau & Słowiński 2004) • The marginal value function ui(xi) y,v,w,zAR Characteristic points of marginal value functions are fixed on actual evaluations of actions from set A

  39. ui(xi) ? ? ? ? ? gi 0 i i yi vi wi zi The UTAGMS method(Greco, Mousseau & Słowiński 2004) • The marginal value function ui(xi) y,v,w,zAR Marginal values in characteristic points are unknown

  40. ui(xi) gi 0 i i yi vi wi zi The UTAGMS method(Greco, Mousseau & Słowiński 2004) • The marginal value function ui(xi) y,v,w,zAR In fact, they are intervals, because all compatible value functions are considered

  41. ui(xi) gi 0 i i yi vi wi zi The UTAGMS method(Greco, Mousseau & Słowiński 2004) • The marginal value function ui(xi) y,v,w,zAR The area of all compatible marginal value functions

  42. ui(xi) gi 0 i i yi vi wi zi The UTAGMS method(Greco, Mousseau & Słowiński 2004) • The marginal value function ui(xi) y,v,w,zAR In the area the marginal compatible value functions must be monotone

  43. ui(xi) gi 0 i i yi vi wi zi The UTAGMS method(Greco, Mousseau & Słowiński 2004) • The marginal value function ui(xi) • This means that the ordinal regression should not seek for m piecewise-linear marginal value functions, but for any compatible additive value function

  44. The UTAGMS method • Let ibe a permutation on the set of actions AR{x,y} that reorders them according to increasing evaluation on criterion gi: where • if AR{x,y}=, then =n+2 • if AR{x,y}={x} or AR{x,y}={y}, then =n+1 • if AR{x,y}={x,y}, then =n • The characteristic points of ui(xi), iI, are then fixen in:

  45. a a b b  a,bAR a a b b E(x,y) a a a The UTAGMS method • For any pair of actions (x,y)A, and for available preference information concerning AR, preference of x over y is determined by compatible value functionsU verifying set E(x,y) of constraints: where  is a small positive constant • For all (x,y)A, E(x,y) = E(y,x)

  46. The UTAGMS method • Nmeans necessary(strong) preference relation • Given a pair of actions x,yA xNyd(x,y)  0 where • d(x,y)  0 means that for allcompatible value functions x is at least as good as y • For x,yAR : xy xNy

  47. The UTAGMS method • Pmeans possible(weak) preference relation • Given a pair of actions x,yA xPyD(x,y)  0 where • D(x,y)  0 means that for at least onecompatible value function x is at least as good as y • For x,yAR : xy notyPx

  48. The UTAGMS method • Some properties: • xNyxPy • Nis a partial preorder (i.e. Nis reflexive and transitive) • Pis strongly complete (i.e. for all x,yA, xPyor yPx)and negatively transitive (i.e. for all x,y,zA, notxPyand not yPznot xPz), (in general, Pis not transitive) • d(x,y) = Min{U(x)–U(y)} = –Max{–[U(x)–U(y)]} = = –Max{U(y)–U(x)} = –D(y,x)

  49. Proof of transitivity of N • d(x,y)>0 means: Min{U(x)-U(y)}>0 • This is equivalent to the fact: for all value functions compatible with the reference preorder, U(x)>U(y) • The set of all compatible value functions is the same for calculation of d(x,y) for any pair x,yA • Suppose, the transitivity of N is not true, i.e. for x,y,zA Min{U(x)-U(y)}>0, Min{U(y)-U(z)}>0, but Min{U(x)-U(z)}<0 • This means that U(x)-U(z) has achieved a minimum value d(x,z)<0 for a value function denoted by U*, such that U*(x)<U*(z), while U*(x)>U*(y) and U*(y)>U*(z) • In other words, U*(x)>U*(y)>U*(z)>U*(x) • This is a contradiction, so N is transitive

  50. The UTAGMS method • Elaboration of the final ranking: • for the necessary preference relation being a partial preorder (N is supported by all compatible value functions) preference: xNy if xNy and not yNx indifference: xNy if xNy and yNx incomparability: x ? y if notxNy and not yNx • for the possible preference relation being complete (P is supported by at least one compatible value function) preference: xPy if xPy and not yPx indifference: xPy if xPy and yPx • N.B. It is impossible to infer one ranking from another because strong and weak outranking relations arenot dual

More Related