Strengthening PROMETHEE with elements from value theory

Σχετικά έγγραφα
ΕΙΣΑΓΩΓΗ ΣΤΗ ΣΤΑΤΙΣΤΙΚΗ ΑΝΑΛΥΣΗ

Math 6 SL Probability Distributions Practice Test Mark Scheme

2 Composition. Invertible Mappings

3.4 SUM AND DIFFERENCE FORMULAS. NOTE: cos(α+β) cos α + cos β cos(α-β) cos α -cos β

ST5224: Advanced Statistical Theory II

5. Choice under Uncertainty

A Bonus-Malus System as a Markov Set-Chain. Małgorzata Niemiec Warsaw School of Economics Institute of Econometrics

Second Order Partial Differential Equations

CHAPTER 25 SOLVING EQUATIONS BY ITERATIVE METHODS

Numerical Analysis FMN011

CRASH COURSE IN PRECALCULUS

Other Test Constructions: Likelihood Ratio & Bayes Tests

Approximation of distance between locations on earth given by latitude and longitude

Practice Exam 2. Conceptual Questions. 1. State a Basic identity and then verify it. (a) Identity: Solution: One identity is csc(θ) = 1

Instruction Execution Times

Partial Differential Equations in Biology The boundary element method. March 26, 2013

Queensland University of Technology Transport Data Analysis and Modeling Methodologies

Biostatistics for Health Sciences Review Sheet

Statistics 104: Quantitative Methods for Economics Formula and Theorem Review

Concrete Mathematics Exercises from 30 September 2016

4.6 Autoregressive Moving Average Model ARMA(1,1)

5.4 The Poisson Distribution.

ΚΥΠΡΙΑΚΗ ΕΤΑΙΡΕΙΑ ΠΛΗΡΟΦΟΡΙΚΗΣ CYPRUS COMPUTER SOCIETY ΠΑΓΚΥΠΡΙΟΣ ΜΑΘΗΤΙΚΟΣ ΔΙΑΓΩΝΙΣΜΟΣ ΠΛΗΡΟΦΟΡΙΚΗΣ 24/3/2007

EE512: Error Control Coding

SCHOOL OF MATHEMATICAL SCIENCES G11LMA Linear Mathematics Examination Solutions

Review Test 3. MULTIPLE CHOICE. Choose the one alternative that best completes the statement or answers the question.

Homework 3 Solutions

ΤΕΧΝΟΛΟΓΙΚΟ ΠΑΝΕΠΙΣΤΗΜΙΟ ΚΥΠΡΟΥ ΣΧΟΛΗ ΓΕΩΤΕΧΝΙΚΩΝ ΕΠΙΣΤΗΜΩΝ ΚΑΙ ΔΙΑΧΕΙΡΙΣΗΣ ΠΕΡΙΒΑΛΛΟΝΤΟΣ. Πτυχιακή εργασία

[1] P Q. Fig. 3.1

Inverse trigonometric functions & General Solution of Trigonometric Equations

Jesse Maassen and Mark Lundstrom Purdue University November 25, 2013

2. THEORY OF EQUATIONS. PREVIOUS EAMCET Bits.

Ψηφιακή Επεξεργασία Εικόνας

k A = [k, k]( )[a 1, a 2 ] = [ka 1,ka 2 ] 4For the division of two intervals of confidence in R +

derivation of the Laplacian from rectangular to spherical coordinates

ΚΥΠΡΙΑΚΗ ΕΤΑΙΡΕΙΑ ΠΛΗΡΟΦΟΡΙΚΗΣ CYPRUS COMPUTER SOCIETY ΠΑΓΚΥΠΡΙΟΣ ΜΑΘΗΤΙΚΟΣ ΔΙΑΓΩΝΙΣΜΟΣ ΠΛΗΡΟΦΟΡΙΚΗΣ 6/5/2006

6.1. Dirac Equation. Hamiltonian. Dirac Eq.

the total number of electrons passing through the lamp.

ΕΛΛΗΝΙΚΗ ΔΗΜΟΚΡΑΤΙΑ ΠΑΝΕΠΙΣΤΗΜΙΟ ΚΡΗΤΗΣ. Ψηφιακή Οικονομία. Διάλεξη 7η: Consumer Behavior Mαρίνα Μπιτσάκη Τμήμα Επιστήμης Υπολογιστών

DESIGN OF MACHINERY SOLUTION MANUAL h in h 4 0.

ω ω ω ω ω ω+2 ω ω+2 + ω ω ω ω+2 + ω ω+1 ω ω+2 2 ω ω ω ω ω ω ω ω+1 ω ω2 ω ω2 + ω ω ω2 + ω ω ω ω2 + ω ω+1 ω ω2 + ω ω+1 + ω ω ω ω2 + ω

Phys460.nb Solution for the t-dependent Schrodinger s equation How did we find the solution? (not required)

Matrices and vectors. Matrix and vector. a 11 a 12 a 1n a 21 a 22 a 2n A = b 1 b 2. b m. R m n, b = = ( a ij. a m1 a m2 a mn. def

PARTIAL NOTES for 6.1 Trigonometric Identities

Ordinal Arithmetic: Addition, Multiplication, Exponentiation and Limit

ΜΕΤΑΠΤΥΧΙΑΚΗ ΔΙΠΛΩΜΑΤΙΚΗ ΕΡΓΑΣΙΑ «ΘΕΜΑ»

6.3 Forecasting ARMA processes

ES440/ES911: CFD. Chapter 5. Solution of Linear Equation Systems

Απόκριση σε Μοναδιαία Ωστική Δύναμη (Unit Impulse) Απόκριση σε Δυνάμεις Αυθαίρετα Μεταβαλλόμενες με το Χρόνο. Απόστολος Σ.

VBA Microsoft Excel. J. Comput. Chem. Jpn., Vol. 5, No. 1, pp (2006)

CHAPTER 48 APPLICATIONS OF MATRICES AND DETERMINANTS

Lecture 34 Bootstrap confidence intervals

상대론적고에너지중이온충돌에서 제트입자와관련된제동복사 박가영 인하대학교 윤진희교수님, 권민정교수님

ΠΟΛΥΤΕΧΝΕΙΟ ΚΡΗΤΗΣ ΣΧΟΛΗ ΜΗΧΑΝΙΚΩΝ ΠΕΡΙΒΑΛΛΟΝΤΟΣ

ΓΕΩΜΕΣΡΙΚΗ ΣΕΚΜΗΡΙΩΗ ΣΟΤ ΙΕΡΟΤ ΝΑΟΤ ΣΟΤ ΣΙΜΙΟΤ ΣΑΤΡΟΤ ΣΟ ΠΕΛΕΝΔΡΙ ΣΗ ΚΤΠΡΟΤ ΜΕ ΕΦΑΡΜΟΓΗ ΑΤΣΟΜΑΣΟΠΟΙΗΜΕΝΟΤ ΤΣΗΜΑΣΟ ΨΗΦΙΑΚΗ ΦΩΣΟΓΡΑΜΜΕΣΡΙΑ

C.S. 430 Assignment 6, Sample Solutions

Matrices and Determinants

(1) Describe the process by which mercury atoms become excited in a fluorescent tube (3)

ΑΛΛΗΛΕΠΙ ΡΑΣΗ ΜΟΡΦΩΝ ΛΥΓΙΣΜΟΥ ΣΤΙΣ ΜΕΤΑΛΛΙΚΕΣ ΚΑΤΑΣΚΕΥΕΣ

The Simply Typed Lambda Calculus

ΚΥΠΡΙΑΚΗ ΕΤΑΙΡΕΙΑ ΠΛΗΡΟΦΟΡΙΚΗΣ CYPRUS COMPUTER SOCIETY ΠΑΓΚΥΠΡΙΟΣ ΜΑΘΗΤΙΚΟΣ ΔΙΑΓΩΝΙΣΜΟΣ ΠΛΗΡΟΦΟΡΙΚΗΣ 19/5/2007

Solutions to the Schrodinger equation atomic orbitals. Ψ 1 s Ψ 2 s Ψ 2 px Ψ 2 py Ψ 2 pz

The challenges of non-stable predicates

Μορφοποίηση υπό όρους : Μορφή > Μορφοποίηση υπό όρους/γραμμές δεδομένων/μορφοποίηση μόο των κελιών που περιέχουν/

Econ 2110: Fall 2008 Suggested Solutions to Problem Set 8 questions or comments to Dan Fetter 1

Biodiesel quality and EN 14214:2012

Mean bond enthalpy Standard enthalpy of formation Bond N H N N N N H O O O

Example Sheet 3 Solutions

CE 530 Molecular Simulation

Figure A.2: MPC and MPCP Age Profiles (estimating ρ, ρ = 2, φ = 0.03)..

ΠΙΣΤΟΠΟΙΗΜΕΝΟ ΕΠΑΓΓΕΛΜΑΤΙΚΟ ΠΕΡΙΓΡΑΜΜΑ «ΧΕΙΡΙΣΤΗΣ ΚΙΝΗΤΩΝ ΜΗΧΑΝΗΜΑΤΩΝ ΜΗΧΑΝΗΜΑΤΩΝ ΕΡΓΟΥ»

Second Order RLC Filters

Μεταπτυχιακή διατριβή. Ανδρέας Παπαευσταθίου

( ) 2 and compare to M.

Chapter 6: Systems of Linear Differential. be continuous functions on the interval

Repeated measures Επαναληπτικές μετρήσεις

Mechanical Behaviour of Materials Chapter 5 Plasticity Theory

Congruence Classes of Invertible Matrices of Order 3 over F 2

Every set of first-order formulas is equivalent to an independent set

ΕΛΛΗΝΙΚΗ ΔΗΜΟΚΡΑΤΙΑ ΠΑΝΕΠΙΣΤΗΜΙΟ ΚΡΗΤΗΣ. Ψηφιακή Οικονομία. Διάλεξη 9η: Basics of Game Theory Mαρίνα Μπιτσάκη Τμήμα Επιστήμης Υπολογιστών

Calculating the propagation delay of coaxial cable

Written Examination. Antennas and Propagation (AA ) April 26, 2017.

Solutions to Exercise Sheet 5

Εργαστήριο Ανάπτυξης Εφαρμογών Βάσεων Δεδομένων. Εξάμηνο 7 ο

ΚΥΠΡΙΑΚΟΣ ΣΥΝΔΕΣΜΟΣ ΠΛΗΡΟΦΟΡΙΚΗΣ CYPRUS COMPUTER SOCIETY 21 ος ΠΑΓΚΥΠΡΙΟΣ ΜΑΘΗΤΙΚΟΣ ΔΙΑΓΩΝΙΣΜΟΣ ΠΛΗΡΟΦΟΡΙΚΗΣ Δεύτερος Γύρος - 30 Μαρτίου 2011

Lecture 2: Dirac notation and a review of linear algebra Read Sakurai chapter 1, Baym chatper 3

SCITECH Volume 13, Issue 2 RESEARCH ORGANISATION Published online: March 29, 2018

Group 30. Contents.

ΠΑΝΕΠΙΣΤΗΜΙΟ ΠΕΙΡΑΙΑ ΤΜΗΜΑ ΝΑΥΤΙΛΙΑΚΩΝ ΣΠΟΥΔΩΝ ΠΡΟΓΡΑΜΜΑ ΜΕΤΑΠΤΥΧΙΑΚΩΝ ΣΠΟΥΔΩΝ ΣΤΗΝ ΝΑΥΤΙΛΙΑ

Lifting Entry (continued)

If we restrict the domain of y = sin x to [ π, π ], the restrict function. y = sin x, π 2 x π 2

Reminders: linear functions

forms This gives Remark 1. How to remember the above formulas: Substituting these into the equation we obtain with

Main source: "Discrete-time systems and computer control" by Α. ΣΚΟΔΡΑΣ ΨΗΦΙΑΚΟΣ ΕΛΕΓΧΟΣ ΔΙΑΛΕΞΗ 4 ΔΙΑΦΑΝΕΙΑ 1

ΤΜΗΜΑ ΦΥΣΙΚΩΝ ΠΟΡΩΝ & ΠΕΡΙΒΑΛΛΟΝΤΟΣ

TMA4115 Matematikk 3

Potential Dividers. 46 minutes. 46 marks. Page 1 of 11

Exercises to Statistics of Material Fatigue No. 5


If we restrict the domain of y = sin x to [ π 2, π 2

Transcript:

2η Συνάντηση Πολυκριτήριας Ανάλυσης Αποφάσεων Ελληνική Εταιρία Επιχειρησιακών Ερευνών Χανιά, 2-22 Οκτωβρίου 2004 Strengthening PROMETHEE with elements from value theory Εμπλουτισμός της μεθόδου PROMETHEE με στοιχεία της Πολυκριτήριας Θεωρίας Χρησιμότητας George Mavrotas*, Danae Diakoulaki Laboratory of Industrial & Energy Economics National Technical University of Athens

Contents PROMETHEE Value theory and partial value functions Combination of PROMETHEE with value theory Numerical example Conclusions

PROMETHEE Outranking method Pairwise comparison of the alternatives Results Promethee I, partial preorder (allow incomparabilities) Promethee II, complete preorder (no incomparabilities)

Brief description of PROMETHEE The difference between two alternatives in one criterion is transformed to [0,] using one of six functional forms (criteria types). Pairwise comparison Square matrix with pairwise multicriteria scores Leaving flow=row sum, Entering flow=column sum, Net flow=leaving flow-entering flow Complete preorder=rating of alternatives according to their net flow

Type of criteria 0 0 q 0 p Type Type 2 Type 3 0 q p 0 q p 0 s Type 4 Type 5 Type 6

PROMETHEE parameters q = indifference threshold (types 2, 4, 5) p = strict preference threshold (types 3, 4, 5) s = gaussian s standard deviation (type 6) Basic assumption: Parameters are constant throughout the entire range of the criterion

Implication The pairwise outranking score p j (a,b) depends only on the difference between the two alternatives (d ab ) and not on their position in the criterion scale. This implies a partial value function (p.v.f.) with a constant slope for the criteria (linear p.v.f.)

Example: Car selection Criterion: maximization of car s speed A 20 km/h C 220 km/h B 50 km/h D 250 km/h B is preferred against A the same as D is preferred against C

Remarks on the example Sometimes the same distance between alternatives is differently perceived, depending on the position of the alternatives in the criterion scale. Example of car s speed criterion: usually we are more sensitive in the differences at the beginning of the scale (normal speed figures) than at the end of the scale (Formula speed figures). In some cases the uniform degree of preference throughout the criterion range is not realistic Remedy: nonlinear partial value functions

Partial value functions (p.v.f.) Three usual types 0 0 0 linear Uniform preference of the DM convex The DM is more sensitive in differences at the end of the scale concave The DM is more sensitive in differences at the beginning of the scale

Car s example revisited The DM is more sensitive in differences between low figures Car speed weak preference strong preference 0 20 50 220 250

Incorporation in PROMETHEE Enrichment of PROMETHEE with elements from value theory Flexible PROMETHEE The DM may choose the shape of the p.v.f. for a criterion, if he/she considers that it is different than the underlying linear.

Preference elicitation Estimate the shape of the partial value function Techniques Bisection (indicate the mid-value point) Difference (compare differences in values)

In practice The DM expresses his/her preferences as in normal PROMETHEE Criterion type ( 6) Required parameters (q, p, s) Weights (w, w 2,, w m ) One additional piece of information: A rough estimate of the partial value function for each criterion (not detailed estimation)

Categories of p.v.f. The DM chooses a representative partial value function for each criterion among five categories: concave more or less concave linear more or less convex convex 0.75 0.5 0.25 concave more or less concave linear more or less convex convex 0

Measuring the position of the compared alternatives: the centrifugal coefficient According to the position of the compared alternatives A and B in the j-th criterion, the centrifugal coefficient is defined as: c jab xaj + xbj midrange j = 2 c [ 0.5, 0.5] range jab j If c > 0 then A and B are considered to be located in the upper half If c < 0 then A and B are considered to be located in the lower half A high c A and B are close to one edge of the criterion scale

Graphical representation range min max A B C D midrange (x A +x B )/2 (x C +x D )/2

Adjustable parameters Assume t is an initially assigned parameter from the DM for the specific criterion (q, p or s). The relative adjusted parameter (t adj ) is calculated using the following relationships: concave: t adj = t (+2c) More or less concave: t adj = t (+c) Linear: t adj = t More or less convex: t adj = t (- c) adjusting coefficents convex: t adj = t (- 2c)

Insight concave: t adj =t (+2c) If c>0 A, B in upper half parameter less sensitive If c<0 A, B in lower half parameter more sensitive Linear Uniform parameter convex: t adj =t(-2c) If c>0 A, B in upper half parameter more sensitive If c<0 A, B in lower half parameter less sensitive

Graphical example Assume a type 5 criterion with a concave partial value function thresholds for two alternatives in the lower half thresholds for two alternatives in the upper half 0 q L q q U p L p p U

Generalization General rule Value of parameter intensity of preference s s 2 0 q q 2 0 Type 2 Type 6 The smaller the parameter the easier we move from indifference to strict preference

Numerical example Problem: choose action plan in order to comply with ozon standards for year 200 in a Greek city 6 candidate action plans (scenarios) 3 criteria Cost (million ) Social Acceptance (qualitative scale) Expert opinion Environmental Safety Margin distance from the imposed limit (20 μg/m 3 ) in terms of concentration (μg/m 3 ).

Evaluation matrix and criteria characteristics AP AP2 Cost (million ) 40 80 Social acceptance 6.4 7.5 Safety margin (μg/m 3 ) 8 2 We are more sensitive in differences near to the imposed limit. AP3 AP4 AP5 AP6 90 50 20 70 7.5 8.6 4.3 6.9 24 28 5 0 The same difference counts more if we are close to the limit than far away from it direction Criteria characteristics min max max Why? type 5 5 5 q p weight p.v.f. shape 5 30 0.4 linear 0.5 0.3 linear 2 5 0.3 concave Motivation for the proposed approach

Example of calculations in third criterion Consider AP with AP2 and AP3 with AP4 Normal Promethee: p j (a,b) = (d ab q)/(p-q) p 3 (ΑP2, AP)=(4-2)/(5-2)=0.67 p 3 (ΑP4, AP3)=(4-2)/(5-2)=0.67

Promethee with adjusted parameters Centrifugal coefficient: c jab x Aj + = 2 x Bj midrange range j j =(8+2)/2 c(ap2,ap)=(0-6.5)/23=-0.28 +2c=0.44 midrange = (28+5)/2 range = (28-5) c(ap4,ap3)=(26-6.5)/23= 0.4 +2c=.82 =(24+28)/2

Graphical representation 2.2 5 9. (=.82*5) (=0.44*5) 0.67 p 3 (AP2, AP)= p 3 (ΑP4, AP3)=0.05 0 The same difference results in diverse pairwise outranking scores 0.88 2 3.6 4 (=0.44*2) (=.82*2)

Results The algorithm was coded in VBA for Excel Flexible Promethee (adjusted parameters) Normal Promethee (uniform parameters) Φ+ Φ- Φ rank Φ+ Φ- Φ rank AP 0.44 0.47-0.03 4 AP 0.40 0.39 0.0 3 AP2 0.39 0.36 0.03 3 AP2 0.33 0.36-0.03 4 AP3 0.45 0.29 0.6 AP3 0.45 0.32 0.3 2 AP4 0.54 0.40 0.4 2 AP4 0.58 0.40 0.8 AP5 0.37 0.60-0.23 6 AP5 0.37 0.56-0.9 6 AP6 0.32 0.40-0.08 5 AP6 0.26 0.36-0.0 5

Comparison Cost (million ) Social acceptance Safety margin (μg/m 3 ) Rank (Normal Promethee) Rank (Flexible Promethee) AP 40 6.4 8 3 4 AP2 80 7.5 2 4 3 AP3 90 7.5 24 2 AP4 50 8.6 28 2 AP5 20 4.3 5 6 6 AP6 70 6.9 0 5 5

Conclusions The idea is to stengthen PROMETHEE with elements from value theory Adjustable parameters offer more flexibility Simple information required by the DM Easily implemented as an extra option in normal PROMETHEE

Thank you for your attention