Changelog

v1.*

v1.0.0

• stabilize API

• rename abstract method Sorter.clusterize to Sorter.sort

• add user API module-level doc-strings

• ensure compatibility with pandas<2.1 while complying with pandas>=2.1

v0.*

v0.7.0

• split implementation from API

  • move all source code into mcda.internal subpackage

  • add API package hierarchy in mcda package

• use user API classes/functions in unit tests

• make graphviz/matplotlib dependencies optional

• complete unit tests for whole package except mcda.plot

• hide all ‘apply’ methods in callables

• mcda.core.functions

  • replace Interval.is_empty method by a property

  • replace PieceWiseFunction.continuous method by a property

• mcda.core.matrices

  • raise KeyError instead of ValueError when IAdjacencyMatrix has different set of labels for rows and columns

  • move to_graph/plot/save_plot methods from IAdjacencyMatrix to AdjacencyValueMatrix

  • raise KeyError if criteria are not unique when building a PartialValueMatrix

  • use new view types for *_values properties to replace dictionaries

• mcda.core.relations

  • hide all classes _validate method

  • make Relation.types method static

  • replace PreferenceStructure get/del methods with multiple accessors and views

• mcda.core.scales

  • make QuantitativeScale.normal method static

  • make DiscreteQuantitativeScale.binary method static

  • replace NormalScale/BinaryScale classes with type aliases

• mcda.core.set_functions

  • hide ISetFunction.validate method

  • add argument to enable/disable ISetFunction object validation on construction (enabled by default)

  • make SetFunction.uniform_capacity method static

• mcda.core.values

  • add new class Value

  • make Values a mappable of Value

v0.6.1

• add AdditivePerformanceTable class which adds aggregated_scale attribute to row-wise sums

• add the Sum aggregator

• AdditiveValueFunctions changes

  • make it a subclass of CriteriaFunctions

  • add an aggregate method which sums values

  • return an AdditivePerformanceTable object when passed on a PerformanceTable

• inverse order of Electre3/ElectreTri constructor arguments preference and indifference thresholds

v0.6.0

• add support for python 3.12

• remove python 3.7 from supported versions

• rework completely API of whole package (only most notable changes listed below)

  • change organizations of scales

  • reorganize main data structures

  • add stype (in_stype/out_stype) attribute to all classes manipulating data

  • change callable classes so typing can be tracked while using their __call__ method (and hide all input type specific methods)

  • add aliases for PreferenceDirection values

  • replace OutrankingMatrix/Ranking classes with type aliases

  • move data transform/normalize methods to mcda.core.transformers module

  • integrate performance table as an attribute of all MCDA algotithms

• change discordance method in electre so it returns partial values

  • change veto thresholds so they are defined per criterion

• add cell property to access IMatrix cells directly

• speed up PreferenceStructure to/from outranking matrix conversion methods

v0.5.1

• remove python 3.6 from supported versions

• make internal classes private

• core

  • matrices

    • remove all block matrix classes

    • forbid duplicated index/column labels

  • performance_tables

    • add concat class method to PerformanceTable

    • replace Container type with simple dict in alternative/criteria values

  • values

    • forbid duplicated index labels

    • remove Container class and replace its usage with simple dict

• outranking

  • transform profiles to performance table scales when applying ElectreTri and SRMP (bugfix)

v0.5.0

• document inherited members

• rework Documentation

• add notebooks to documentation

• core

  • values:

    • add strong and normal dominance methods to ScaleValues

    • add generic Container class

    • rename transform_to method of ScaleValues to transform

  • performance_table

    • add efficients property to PerformanceTable

    • merge apply_ methods in one methos apply

    • replace get_ methods by [alternatives/criteria]_values container views

  • matrices

    • add classes for block matrix and profile Vs alternatives matrix

    • add combinations of classes in matrix module

  • make all __repr__ methods uncovered by unit tests

  • add categories module

  • add criteria_functions module

• outranking

  • fix wrong behaviour of ElectreTri class

  • change type of concordance, discordance and credibility matrices in Electre

  • reorganize electre class hierarchy

  • reorganize promethee module with a more modular structure

v0.4.1

• outranking: fix bug in ElectreTri which failed when any other set of criteria than 0..N was used

v0.4.0

• replace API by an object oriented one in whole package

• add preference elicitation algorithm for SRMP

v0.3.6

• fix SRMP progressive ranking plot function

v0.3.5

• replace all criteria/alternatives values by labelled dict, pandas Series or DataFrames

• move all matrix related functions needed by mcda.core.relations into it

• core

  • replace list of lists by pandas dataframe for performance table

  • split sorting module in relations module

  • relations: add basic relations functions

• outranking

  • add SRMP algorithm

  • remove custom graph data structure

• plot

  • change plot_relation so it can plot any relations without any supposition

  • add area plot, parallel coordinates plot, stacked bar plot

  • add text and annotations

  • add horizontal and vertical stripes

  • add legend

  • add color and alpha parameters

  • add linestyle and marker style parameters for line plot

v0.3.4

• core

  • rework fuzzy scales and numbers (trapezoidal shape only)

• mavt

  • add OWA divergence and balance measures

  • add ULOWA aggregator

• outranking

  • add promethee1 algorithm

  • add Promethee 2 algorithm

  • add Electre 2 algorithm

  • add Electre 3 algorithm

  • add Electre TRI Algorithm

  • add utils.py for management functions

• plot

  • plot promethee GAIA

  • plot method for graph using matrix or relations_list

  • matrix plot can select displayed values

v0.3.3

• mavt

  • add OWA aggregator and related functions

• outranking

  • add Electre 1 algorithm

  • plot method for graph

v0.3.2

• mavt

  • show all aggregators functions in the API

v0.3.1

• plot

  • add rlimits attribute for RadarPlot

v0.3.0

• core

  • add sets and set_functions modules

• mavt

  • add choquet integral aggregators

• set public API for all modules (__all__ module constants)

• change function names

  • core.performance_table.valid_alternatives -> in_scales_alternatives

  • mavt.aggregators.weighted_sum_after_normalization -> normalized_weighted_sum

  • mavt.uta.generate_g_matrix -> generate_criteria_values_matrix

  • mavt.uta.generate_u_variables -> generate_marginal_utility_variables

  • mavt.uta.generate_sigma_variables -> generate_alternatives_errors_variables

  • mavt.uta.compute_uta_after_normalization -> normalized_uta

  • mavt.uta.compute_uta_star_after_normalization -> normalized_uta_star

  • mavt.uta.compute_post_optimality -> post_optimality

  • mavt.uta.compute_star_post_optimality -> star_post_optimality

  • mavt.uta.compute_uta -> uta

  • mavt.uta.compute_uta_star -> uta_star

• remove Matrix type alias and replace all references with List[List[]]

• rework completely plotting utilities in subpackage plot

v0.2.1

• fix mypy errors when importing this package

v0.2.0

• add py.typed file to packages files for allowing type-checking on import

• mavt

  • add weighted sum aggregator

  • add UTA & UTA* algorithm

• core.performance_table

  • add function to find alternatives with performances in scales

  • add subtable function

• core.sorting

  • add relation type enum

• generalize transform & normalize & range methods

• add functions with an input scale that can be transformed

• rename nominative scale to nominal scale

v0.1.0

(start of the changelog)