Creating final rankings using MILP method should return all possible rankings. So it is necessary to run the creation process multiple times, adding after each proper restriction.

Hint: it could be done like finding inconsistency in UTA method

Right now you need to create an instance of proper class in order to visualize graphs or tables. It should be done automatically from Outranking (outranking.outranking) class.

Find a good package and use that tool in development mode.

• Add dev dependency
• Format all project files
• Add checking / auto-formatting process into gh-actions pipeline

Because of different calculations, it is necessary to split in some way ranking creation for ELECTRE and PROMETHEE type matrices.

• Find proper design pattern for Valued Outranking class
• Implement calculations for both methods

Create another one visualization feature which should show differences between source outranking matrix and final outranking matrix. It can be done using matplotlib.

Provide new experiments directory, where research will be made. At first add methods for randomly generated matrices.

Change Outranking class in such way which will save some lines of code, because for each method (Crisp, Valued and StochasticAcceptance) there are similar calculation processes.

Change way of drawing final rankings, adding restriction of place variant boxes in proper level (y-axis). It could be hard, because of using graphviz library.

If necessary, it could be done, changing the library to networkX