promethee

This module implements Promethee algorithms in a heavily modular way.

Todo

Implement commented-out classes

class GeneralizedCriteria(generalized_criterions)

Bases: object

This class defines generalized criteria.

Parameters

generalized_criterions (Mapping[Any, GeneralizedCriterion]) –

__call__(x: Series, y: Series, /, *, preference_directions: Optional[Mapping[Any, Optional[PreferenceDirection]]] = None) → Series

__call__(x: DataFrame, /, *, preference_directions: Optional[Mapping[Any, Optional[PreferenceDirection]]] = None) → DataFrame

__call__(x: Values[QuantitativeScale], y: Values[QuantitativeScale], /) → Values[QuantitativeScale]

__call__(x: PerformanceTable[QuantitativeScale], /) → PartialValueMatrix[QuantitativeScale]

Apply generalized criteria to inputs.

Parameters

• x – first input

• y – second input (only if comparing one pair of alternative values)

• preference_directions –

Raises

TypeError – if arguments don’t match one of the implemented apply_ method

Returns

result

within_in_scales(data)

Check potential input is inside input scales.

Parameters

data (Union[Series, Values[QuantitativeScale], DataFrame, PerformanceTable[QuantitativeScale]]) –

Returns

True if data is within scales

Return type

bool

class GeneralizedCriterion(preference_function, scale=None)

Bases: object

This class defines a generalized criterion which is used to compare alternatives according to a criterion.

Implementations naming conventions are taken from [FGE05].

Parameters

• preference_function (NumericFunction) – function applied on alternative values difference

• scale (Union[QuantitativeScale, None]) – input scale (inferred from input if not provided)

Note

Alternatives values comparisons are transformed so that a positive difference implies the first alternative is preferred. preference_function is then applied on this difference.

Warning

scale is only declarative, and not use for checking/transforming data.

__call__(x: float, y: float, /, *, preference_direction: Optional[PreferenceDirection] = None) → float

__call__(x: Series, /, *, preference_direction: Optional[PreferenceDirection] = None) → DataFrame

__call__(x: CommensurableValues[QuantitativeScale], /) → AdjacencyValueMatrix[QuantitativeScale]

Apply generalized criterion to inputs.

Parameters

• x – first input

• y – second input (only if comparing two criterion values)

• preference_direction –

Raises

TypeError – if arguments don’t match one of the implemented apply_ method

Returns

result

within_scale(data)

Check potential input is inside input scale.

Parameters

data (Union[float, Series, CommensurableValues[QuantitativeScale]]) –

Returns

True if data is within scale

Return type

bool

class Promethee(performance_table, preference_functions, generalized_criteria_aggregator)

Bases: ABC

This class describes the common interface for all Promethee algorithms.

Parameters

• performance_table (PerformanceTable[QuantitativeScale]) –

• preference_functions (Mapping[Any, NumericFunction]) –

• generalized_criteria_aggregator (Aggregator[QuantitativeScale, QuantitativeScale]) –

Raises

TypeError – if performance_table is not ordinal

abstract flows(preferences, **kwargs)

Compute flows.

Parameters

preferences (AdjacencyValueMatrix[QuantitativeScale]) –

Returns

flows

Return type

Series

partial_preferences()

Compute partial preferences for each alternatives’ pair.

Parameters

performance_table –

Returns

partial preferences

Return type

PartialValueMatrix[QuantitativeScale]

preferences(partial_preferences)

Compute preferences for each alternatives’ pair from partial preferences.

Parameters

partial_preferences (PartialValueMatrix[QuantitativeScale]) –

Returns

preferences

Return type

AdjacencyValueMatrix[QuantitativeScale]

criteria_flows(partial_preferences)

Returns the criteria flows.

Parameters

partial_preferences (PartialValueMatrix[QuantitativeScale]) –

Returns

Return type

DataFrame

negative_flows(preferences, profiles=None)

Compute negative flows.

Parameters

• preferences (AdjacencyValueMatrix[QuantitativeScale]) –

• profiles (Optional[List]) –

Returns

computed negative flows

Return type

Series

Note

if profiles is not None, it will returns profiles vs alternatives negative flows

net_outranking_flows(preferences)

Compute net outranking flows.

Parameters

preferences (AdjacencyValueMatrix[QuantitativeScale]) –

Returns

net outranking flows

Return type

Series

positive_flows(preferences, profiles=None)

Compute positive flows.

Parameters

• preferences (AdjacencyValueMatrix[QuantitativeScale]) –

• profiles (Optional[List]) –

Returns

computed positive flows

Return type

Series

Note

if profiles is not None, it will returns alternatives vs profiles positive flows