RORUTA: necessary and possible relations

Use of the widgets RORUTA-NecessaryAndPossiblePreferenceRelations and RORUTA-RepresentativeValueFunction from diviz.

Dependencies

• Python3; if you want to plot the most representative utility function, the matplotlib library is needed also
• diviz (tested on the version 1.15.1)

Useful notes

One should proceed as follows:

1. Use the script roruta.py to create all input files and include them into the workflow in diviz.
2. Run the workflow.
3. Use the script roruta.py to analyse the results:
   • draw necessary and possible relations,
   • plot the most representative utility function,
   • evaluate the most representative utility function on the alternatives.

First of all, the following values must be set:

• divizWFfolder: the chosen workflow folder, where diviz outputs are stored
• performanceTableCSV: name of the file, where a performance table is stored. The file can be the result of a call of the function createCSVPerformanceTable, whose documentation includes also the expected form of the performance table in this file. If the number of alternatives or criteria is high, it might me tedious to enter the values of the table one by one (which is the case, when we use createCSVPerformanceTable). In that case, user should find a better way to create a table.
• inputFolder: location of performanceTableCSV and folders with settings files for the diviz RORUTA-related project projectName
• myProjects: folder in inputFolder, where folders with setings file are stored
• projectName: the name of a diviz workflow that is used to produce the results. In the folder inputFolder/projectName, all .xml settings file are created by the functions, such as alternativesXML etc. In the folder divizWFfolder/projectName, one can find all files, produced by diviz, when running the workflow.
• preferences: this folder is not necessary, if we get user-defined preferences from some other source (for example, by defining them directly in roruta.py) or if we simply do not have any. However, if the folder is there, we expect that it contains .pref files which are ordinary text files, that must be compatible with the function that reads them: defineStrongRelations, hence its documentations also includes the form of the .pref files.

Strings describing paths should contain / and not \, and should not end with / (function toSlash might come in handy).

The following file structure is expected:

Input files

inputFolder/
  performanceTableCSV (file)
  myProjects          (folder)
      projectName     (folder)
      ....
  preferences         (folder)
  ...

Output files

divizWFfolder/
    projectName/
        standard structure of output (and input) files and folders, produced by diviz

It is necessary for the diviz workflow with the name projectName, to have the following widgets:

• RORUTA-NecessaryAndPossiblePreferenceRelations (for the computation of relations),
• RORUTA-RepresentativeValueFunction (for the most representative utility function),

see the .

Defining relations

When defining your relations/preferences, do not use the real names in the pairs: A list of (for example, strog) relations [['Mazda CX-5 ... 2015 - 2016', 'Audi A3 ... 2014 - 2016'], ...] is not valid. You must use the abbreviations of the form a<id> instead of the real names, where 0 <= id < number of alternatives. The id of an alternative is the index of the alternative in the list of alternatives myAlternatives.

The correct version of the upper list with the relations would be [['a1', 'a0'], ...] (since Mazda and Audi are the second and the first car in myAlternatives).

The same holds for weak and indif relations (and for the lists, describing the intensities of the relations). Some examples can be seen in the comments below.

Element [aI, aJ] of the list for relation R (strong: >, weak: >=, indifferent: =), means that aI R aJ. Element [[aI1, aJ1], [aI2, aJ2]] of the list for intensities of the relation R, means that U(aI1) - U(aJ1) R U(aI2) - U(aJ2) for all utility functions U, and R as in the upper case.

Working example

Working example bases on data from this page and the weighted sum model carsExample/weightedSumModel.xlsx.