skcriteria.agg.ervd module

Implementation of ERVD method.

skcriteria.agg.ervd.ervd(matrix, objectives, weights, reference_points, alpha, lambd, metric, w_metric, **kwargs)[source]

Execute ERVD without any validation.

class skcriteria.agg.ervd.ERVD(*, lambda_value=2.25, alpha_value=0.88, metric=functools.partial(<function minkowski>, p=1), w_metric=True)[source]

Bases: SKCDecisionMakerABC

Election based on Relative Value Distances (ERVD) decision-making method.

This method integrates an s-shape value function, departing from the traditional expected utility function, to more accurately capture risk-averse and risk-seeking behaviors. ERVD builds upon the foundational principles of the TOPSIS method, extending its capabilities by incorporating concepts from prospect theory to refine the assessment of alternatives based on their relative distances from ideal and anti-ideal solutions.

Parameters:

  • lambda_value (float, default=2.25) – Represents the attenuation factor of the losse.

  • alpha_value (float, default=0.88) – Diminishing sensitivity parameters.

  • metric (str or callable, default='minkowski') – The distance metric to be used for calculating distances between alternatives and ideal/anti-ideal points. It can be a string representing a metric name from scipy.spatial.distance or a custom callable function that computes distances.

  • w_metric (bool, default=True) – Whether to use weights in the distance metric calculation. If True, the weights will be applied to the alternatives when calculating distances. If False, the distances will be calculated without weights.

References

[Shyur et al., 2015]

property alpha_value

Diminishing sensitivity parameter.

property lambda_value

Attenuation factor of the losses.

property metric

Which distance metric will be used.

property w_metric

Whether to use weights in the metric.

evaluate(dm, *, reference_points=None)[source]

Validate the dm and calculate and evaluate the alternatives.

Parameters:

  • dm (skcriteria.data.DecisionMatrix) – Decision matrix on which the ranking will be calculated.

  • reference_points (array-like, optional) – Reference points for each criterion.

Returns:

Ranking.

Return type:

skcriteria.data.RankResult