MCDA (version 0.1.0)

SRMPInferenceNoInconsist: Exact inference of an SRMP model given a maximum number of reference profiles - no inconsistencies

Description

Exact inference approach from pairwise comparisons of alternatives for the SRMP ranking model. This method only outputs a result when an SRMP model consistent with the provided pairwise comparisons exists. The method will search for a model with the minimum possible number of profiles up to a given maximum value. If such a model exists, this method is significantly faster than the one which handles inconsistencies.

Usage

SRMPInferenceNoInconsist(
  performanceTable,
  criteriaMinMax,
  maxProfilesNumber,
  preferencePairs,
  indifferencePairs = NULL,
  alternativesIDs = NULL,
  criteriaIDs = NULL,
  timeLimit = NULL
)

Value

The function returns a list containing:

criteriaWeights

The inferred criteria weights.

referenceProfilesNumber

The inferred reference profiles number.

referenceProfiles

The inferred reference profiles.

lexicographicOrder

The inferred lexicographic order of the profiles.

solverStatus

The solver status as given by glpk.

humanReadableStatus

A description of the solver status.

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

A strictly pozitive numerical value which gives the highest number of reference profiles the sought SRMP model should have.

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. By default NULL (which corresponds to no time limit).

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

# \donttest{
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)

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)

result<-SRMPInferenceNoInconsist(performanceTable, criteriaMinMax, 3, preferencePairs,
                                 indifferencePairs, alternativesIDs = c("a1","a2","a3","a4",
                                 "a5","a6","a7","a8","a10","a11","a12","a14","a16","a17","a18",
                                 "a19","a20","a21","a23","a24"))
# }