Agile MCDA ModellingXMCDA meets D4

R. Bisdorff & M. Zamfiroiu

Manchester, April 15, 2011Cost Action IC0206 Algorithmic Decision Theory

1

Content

1. The D4 Rubis project

2. Use cases of XMCDA in D4

3. Granularity of the MCDA models

4. UMCDA-ML profiles and stereotypes

5. New horizons for XMCDA development

22

1. The D4 Rubis project

3

4

D4 Rubis project

Rubis applications

Application users

User role

5

User raymondConnects to

D4

6

User selectshis application

Starting the D4Selected

application

7

main steps of the decision aid

process8

a. Edit the set of potential alternatives

Grid fragment

9

b. Edit the family of criteria

10

c. Edit the performances of the actions on the criteria

Filtered performance

11

Declarativedefinition of the tabs

declarative definition of filtering conditions

12

View the entire performance table

use D4 Pythonscripting

13

d. Tuning the preference discrimination

declarative definition ofthresholds

14

15

e. Computing the bipolar valued outranking digraphe. Computing the bipolar valued outranking digraph

16

17

f. Get Rubis best choice recommendation from theRubis Solver web service

18

2. Use cases of XMCDA in D4

19

D4

Rubisclass

instances

D4

Rubisclass

instances

XMCDARubis

Problem

jythonjython

a. Interfacing D4 and digraphs class instances

digraphsclass

instances

digraphsclass

instances

20

Example: showing the pairwise comparison table

D4 classdefinition

class operation

The class instance has a pointer to an XMCDA encoded instance of a complete

Rubis PerformanceTableau

21

D4

class instances

D4

class instances

XMCDARubis

Problem/Solution

RubisWeb

Solver

RubisWeb

Solver

http

Rest

jython

b. Interfacing D4 and the Rubis XMCDA Solver

22

D4 python xmlrpcexchange

with the Rubis Solver

D4 classdefinition

class operation

23

24

UMCDAML perspectives

Learning from the D4 experience

25

Motivation

• D4 ?

• Power-User design tool for MCDA problems

• No deployment cycle : app online from the begining

• Fast : new class in minutes

• Flexible : refactoring with instant impact to data

• Only enrolling you ?

• Technological war ?

26

Motivation

…

27

DesktopApp (d2)

WinApp (IRIS)

DesktopApp/Sv (Diviz)

WebApp (d3)WebAppFactory (d4)

WinApp (Excel)

U ML

High stakes : raise abstraction

Technology vs methodology

Modeling MCDA problems with UML

..

28

Use Cases

Actor

Multi platform, multi language …

…

29Workflow (activity diagram)

Universal visual value

…

30

Customized entities (class diagram)

Object’s lifecycle : workflow regulator

…

31

StateMachine

Can be reflectedin the GUI

3. Granularity of the MCDA models

32

UML granularity concepts

• Model

• Static : package*, class*, component*

• Dynamic : usecase, activity*, state*, operation*

• Metamodel solution

• (*) : Composite design pattern

• XMI : XML Model Interachange

33

© KarmicSoft 2oo6-2oo8

34

…

3. UMCDA-ML profiles and stereotypes

35

UML Profile

• Profile

• provides a generic extension mechanism for customizing UML models for particular domains and platforms.

• define new concepts (meta) called << stereotypes >>more than types, less than metatypes

• model remain compliant with the standard

• finally, the standard can evolve and integrate the stereotype as an official new concept

36

3. New horizons

for XMCDA ?

37

XMCDA

• UML-like profiles and stereotypes

• Needs a tool for diagrams and generation

• <<problem>>, <<package>>,

• <<phase>>, <<workflow>>

• Types aggregates : java-like generics

• Using Spring parsing features

• Set<Alternative>, List<Evaluator> …

• Matrix<Alternative, Evaluator, Critera> …38

Vision

…

39

DesktopApp (d2)

WinApp (IRIS)

DesktopApp/Sv (Diviz)

WebApp (d3)WebAppFactory (d4)

WinApp (Excel)

U ML

Methdological power

Technological freedom

Strong methodological concepts +

Rich exhange mechanisms

(xMCDA + WebServices)=> Technological diversity

becomes an asset

Q&A

4040