SRMPInferenceApprox: Approximative inference of an SRMP model

In MCDA: Support for the Multicriteria Decision Aiding Process

Approximative inference of an SRMP model

Description

Approximative inference approach from pairwise comparisons of alternatives for the SRMP ranking model. This method outputs an SRMP model that fulfils as many pairwise comparisons as possible. Neither the number of reference profiles, nor the lexicographic order are fixed beforehand, however a maximum value for the number of reference profiles needs to be provided.

Usage

SRMPInferenceApprox(
  performanceTable,
  criteriaMinMax,
  maxProfilesNumber,
  preferencePairs,
  indifferencePairs = NULL,
  alternativesIDs = NULL,
  criteriaIDs = NULL,
  timeLimit = 60,
  populationSize = 20,
  mutationProb = 0.1
)

Arguments

performanceTable	Matrix or data frame containing the performance table. Each row corresponds to an alternative, and each column to a criterion. Rows (resp. columns) must be named according to the IDs of the alternatives (resp. criteria).
criteriaMinMax	Vector containing the preference direction on each of the criteria. "min" (resp. "max") indicates that the criterion has to be minimized (maximized). The elements are named according to the IDs of the criteria.
maxProfilesNumber	The maximum number of reference profiles of the SRMP model.
preferencePairs	A two column matrix containing on each row a pair of alternative names where the first alternative is considered to be strictly preferred to the second.
indifferencePairs	A two column matrix containing on each row a pair of alternative names the two alternatives are considered to indifferent with respect to each other.
alternativesIDs	Vector containing IDs of alternatives, according to which the datashould be filtered.
criteriaIDs	Vector containing IDs of criteria, according to which the data should be filtered.
timeLimit	Allows to fix a time limit of the execution, in seconds (default 60).
populationSize	Allows to change the size of the population used by the genetic algorithm (default 20).
mutationProb	Allows to change the mutation probability used by the genetic algorithm (default 0.1).

Value

The function returns a list containing:

criteriaWeights	The inferred criteria weights.
referenceProfilesNumber	The number of inferred reference profiles.
referenceProfiles	The inferred reference profiles.
lexicographicOrder	The inferred lexicographic order of the reference profiles.
fitness	The percentage of fulfilled pair-wise relations.

References

A-L. OLTEANU, V. MOUSSEAU, W. OUERDANE, A. ROLLAND, Y. ZHENG, Preference Elicitation for a Ranking Method based on Multiple Reference Profiles, forthcoming 2018.

Examples

performanceTable <- rbind(c(10,10,9),c(10,9,10),c(9,10,10),c(9,9,10),c(9,10,9),c(10,9,9),
                          c(10,10,7),c(10,7,10),c(7,10,10),c(9,9,17),c(9,17,9),c(17,9,9),
                          c(7,10,17),c(10,17,7),c(17,7,10),c(7,17,10),c(17,10,7),c(10,7,17),
                          c(7,9,17),c(9,17,7),c(17,7,9),c(7,17,9),c(17,9,7),c(9,7,17))

criteriaMinMax <- c("max","max","max")

rownames(performanceTable) <- c("a1","a2","a3","a4","a5","a6","a7","a8","a9","a10","a11",
                                "a12","a13","a14","a15","a16","a17","a18","a19","a20",
                                "a21","a22","a23","a24")

colnames(performanceTable) <- c("c1","c2","c3")

names(criteriaMinMax) <- colnames(performanceTable)

# expected result for the tests below

expectedpreorder <- list("a16","a13",c("a3","a9"),"a14","a17",c("a1","a7"),"a18","a15")

# test - preferences and indifferences

preferencePairs <- matrix(c("a16","a13","a3","a14","a17","a1","a18","a15","a2","a11",
                            "a5","a10","a4","a12","a13","a3","a14","a17","a1","a18",
                            "a15","a2","a11","a5","a10","a4","a12","a6"),14,2)
indifferencePairs <- matrix(c("a3","a1","a2","a11","a11","a20","a10","a10","a19","a12",
                              "a12","a21","a9","a7","a8","a20","a22","a22","a19","a24",
                              "a24","a21","a23","a23"),12,2)

set.seed(1)

result<-SRMPInferenceApprox(performanceTable, criteriaMinMax, 3, preferencePairs,
                            indifferencePairs, alternativesIDs = c("a1","a3","a7",
                            "a9","a13","a14","a15","a16","a17","a18"))

MCDA documentation built on Nov. 24, 2023, 5:10 p.m.