PaperlayerView exact revision on arXiv

Source-linked AI summary

Recommending Multiple Criteria Decision Analysis Methods with A New Taxonomy-based Decision Support System

Marco Cinelli, Miłosz Kadziński, Grzegorz Miebs, Michael Gonzalez, Roman Słowiński

arXiv:2106.07378v1cs.AIcs.OH

TL;DR

Selecting an appropriate MCDA method is complicated by the many characteristics of decision-making problems. The MCDA-MSS addresses this with a taxonomy-based decision support system whose taxonomy completely described the case studies tested.

  • Problem

    MCDA processes involve several complexities, making the selection of an appropriate method a recurring challenge.

  • Method

    The paper proposes a decision support system for selecting MCDA methods using a taxonomy and database organized around decision-problem characteristics and desired recommendations.

  • Results

    The MCDA-MSS taxonomy completely described all the case studies examined in the reported application.

  • Takeaways & Limitations

    The MCDA-MSS supports method-selection analysis for complex MCDA processes and can identify methodological mistakes associated with selecting a single-score model.

  • Takeaways & Limitations

    The database contains 31,980 manually assigned binary values, so other researchers may disagree with some mappings of MCDA methods.

Abstract

from arXiv · show

We present the Multiple Criteria Decision Analysis Methods Selection Software (MCDA-MSS). This decision support system helps analysts answering a recurring question in decision science: Which is the most suitable Multiple Criteria Decision Analysis method (or a subset of MCDA methods) that should be used for a given Decision-Making Problem (DMP)?. The MCDA-MSS includes guidance to lead decision-making processes and choose among an extensive collection (over 200) of MCDA methods. These are assessed according to an original comprehensive set of problem characteristics. The accounted features concern problem formulation, preference elicitation and types of preference information, desired features of a preference model, and construction of the decision recommendation. The applicability of the MCDA-MSS has been tested on several case studies. The MCDA-MSS includes the capabilities of (i) covering from very simple to very complex DMPs, (ii) offering recommendations for DMPs that do not match any method from the collection, (iii) helping analysts prioritize efforts for reducing gaps in the description of the DMPs, and (iv) unveiling methodological mistakes that occur in the selection of the methods. A community-wide initiative involving experts in MCDA methodology, analysts using these methods, and decision-makers receiving decision recommendations will contribute to expansion of the MCDA-MSS.

3 Systems Research Institute, Polish Academy of Sciences, Newelska 6, 01-447 Warsaw, Poland

The supplied passage identifies the paper and its authors.

  • The paper is titled “Multiple Criteria Decision Analysis Methods Selection Software (MCDA-MSS).”It is attributed to Marco Cinelli, Miłosz Kadziński, Grzegorz Miebs, Michael Gonzalez, and Roman Słowiński.

1. Introduction

The paper addresses the difficulty of selecting an appropriate MCDA method for a given decision-making problem. It introduces MCDA-MSS, a taxonomy-based decision support system intended to support structured method selection across diverse and complex problems.

  • 1. Introduction: MCDA-MSS recommends suitable MCDA methods for a given decision-making problem by matching problem characteristics to method capabilities.The matching process can seek methods that satisfy all or as many specified characteristics as possible.
  • 1. Introduction: The system targets a recurring selection dilemma created by the steadily growing number of MCDA methods.Analysts must account for the decision problem’s formulation, alternatives, criteria, preferences, and desired recommendation.
  • 1. Introduction: Inadequate method selection can neglect critical problem aspects, produce undesired compromises, and yield recommendations misaligned with the problem’s assessment of alternatives.The paper therefore motivates tools that assist analysts in choosing methods for specific applications.
  • 1.3. Main contributions of the paper: MCDA-MSS uses a comprehensive taxonomy of 156 characteristics and a library containing more than 200 MCDA methods.The taxonomy expands an earlier 66-characteristic vocabulary and covers diverse methodological approaches.
  • 1.3. Main contributions of the paper: The system supports unmatched problems through dialogue with decision-makers, prioritizes uncertainty reduction, and helps identify methodological selection errors.These capabilities are presented as contributions beyond existing decision-support systems.

2. MCDA-MSS development

The MCDA-MSS was developed through a three-stage methodology. The stages covered taxonomy construction, software development, and testing on case studies.

  • 2. MCDA-MSS development: The development methodology comprised three stages: taxonomy design, web-software development, and case-study testing.The final stage assessed the system’s usability and performance.
  • 2. MCDA-MSS development: The first stage shaped the taxonomy used to describe each MCDA method in the system’s library.
  • 2. MCDA-MSS development: The second stage developed the MCDA-MSS in web-software form, while the third tested it on a series of case studies.

2.1 Stage 1: Develop the taxonomy and the database of MCDA methods used in the MCDA-MSS

Stage 1 developed taxonomy v.2 and an initially wide database of MCDA methods for the MCDA-MSS. The taxonomy describes decision problems through four sections, while the database contains 205 mapped methods and focuses currently on single-decision-maker methods.

  • The current method database focuses on single-decision-maker methods, while methods for group decision-makers remain a possible future inclusion.
  • Taxonomy v.2 expanded and restructured earlier taxonomy phases while excluding qualitative features such as ease of use and data-collection time.
  • The taxonomy contains 156 objective features in version 2, compared with 66 in taxonomy v.1.
  • The taxonomy has four sections covering problem typology, preference models, preference elicitation, and exploitation of the preference model.
  • 205 MCDA methods are included in the MCDA-MSS database and mapped according to the taxonomy.The database was intended as an initially wide list that can be complemented in the future.

2.2 Stage 2: Development of the MCDA-MSS in a web-software

The MCDA-MSS uses taxonomy-based rule modeling to match decision-problem characteristics with suitable MCDA methods. Its web interface structures method information, explains questions, handles incomplete matches, and helps analysts reduce uncertainty efficiently.

  • Rule-based modeling: The MCDA-MSS applies rule-based modeling in which problem characteristics activate recommendations for suitable MCDA methods.The rules map conjunctions of activated or satisfied characteristics to one or more recommended methods.
  • Interactive structure: The software organizes method information into four sections and presents it through a sequenced, stepwise questioning process.Information boxes explain each question and transfer knowledge to decision-makers during the interaction.
  • Method repository: The MCDA-MSS contains more than 200 established and recent MCDA methods organized in a taxonomy-based database.The database maps methods to subsets of taxonomy features for different problem statements.
  • Incomplete matches: When no method fully matches a decision problem, the recommender identifies methods that satisfy selected binding features and displays their missed features.This supports recommendations that are as close as possible while making the mismatch explicit.
  • Incomplete matches: Partially matching methods should not be applied blindly; analysts must discuss missed features and may reformulate the problem or develop a tailored method.The paper identifies problem re-discussion and development of a new or tailored method as possible courses of action.
  • Question prioritization: The software identifies the most selective unanswered questions so analysts can prioritize interactions that reduce the recommended method set most.This capability helps focus discussions with the decision-maker on questions expected to provide the greatest reduction.

3. The MCDA-MSS in action

The MCDA-MSS was tested on nine literature case studies to assess applicability across real-life decision problems and compare selected methods with its recommendations. The test found both matching recommendations and mismatches that exposed missed decision-making features and possible method-selection errors.

  • 3.1 Set-up for the MCDA-MSS test with literature case studies: Nine literature case studies tested whether MCDA-MSS recommendations matched methods previously chosen by analysts.The cases covered applications including airline service quality, land remediation, biofuels, urban regeneration, supplier selection, land-use suitability, and energy planning.
  • 3.1 Set-up for the MCDA-MSS test with literature case studies: The case-study evaluation was preliminary and was intended to assess transparent guidance through method selection rather than establish general applicability or performance conclusions.The authors did not aim to assemble a large set of studies from one application area for broad conclusions.
  • 3.1 Set-up for the MCDA-MSS test with literature case studies: The taxonomy described all case studies, while Table 5 compared authors’ methods, MCDA-MSS recommendations, and missed decision-making features.The software represents taxonomy features as questions that describe both MCDA methods and decision-making problems, enabling direct matching.
  • 3.2 Results of the application of the MCDA-MSS to literature case studies: Four case studies produced mismatches because the selected methods were not among those recommended, allowing MCDA-MSS to flag possible method-selection errors.The mismatch analysis attributes discrepancies to differences between the features supported by the selected method and those required by the case study.
  • 3.2 Results of the application of the MCDA-MSS to literature case studies: The software identified mismatches involving ordinal or qualitative scales, inappropriate weighting, missing criterion interactions, and ranking methods applied to sorting problems.Examples include weighted additive methods used for sorting problems, qualitative criteria numerically coded as quantitative, and omitted interactions such as price-related criteria.
  • 3.2 Results of the application of the MCDA-MSS to literature case studies: For sorting problems, MCDA-MSS recommended sorting methods such as FlowSort instead of ranking methods that could not explicitly assign priority levels.One case involved an increasing set of alternatives and an incomplete set of criteria, prompting recommendations that missed only the criteria-set feature and requiring caution in interpreting recommendations.

4. Conclusions

The MCDA-MSS structures the description of decision-making problems and recommends suitable MCDA methods, including when no method completely matches. Tests and proposed extensions support transparent, accountable, and continually expanding method selection.

  • Key contributions: The MCDA-MSS describes complex decision-making problems and links their characteristics to suitable MCDA methods through a structured, stepwise procedure.Its database contains 205 methods and 156 characteristics, covering problems from very simple to very complex.
  • Key contributions: It recommends methods even when no method perfectly matches the decision-making problem, using the available problem information and method knowledge.This prevents the analyst from being left without guidance when the method collection contains no complete match.
  • Key contributions: The system prioritizes unanswered questions that most efficiently reduce the remaining method options when the decision problem cannot be fully described.This refinement strategy directs analysts toward the most discriminatory questions.
  • Key contributions: The MCDA-MSS can expose methodological mistakes, including treating weights as importance coefficients rather than trade-offs or using ordinal scales where quantitative scales are required.The authors state that identifying these mismatches can help avoid similar mistakes in future case studies.
  • Testing and future research: Testing on ten literature case studies confirmed that the MCDA-MSS can describe varied real-life problems, identify suitable methods, and reveal flaws in method selection.The system also creates a recordable description of the choices leading to method selection, supporting consistent communication and accountability.
  • Testing and future research: Future work includes expanding the method repository and case-study tests, adding decision-aiding features, and incorporating methods for multiple decision-makers.The proposed community initiative involves MCDA experts, analysts, and decision-makers.

5. Disclaimer

The disclaimer states that the research reflects the authors’ opinions rather than necessarily those of the U.S. Environmental Protection Agency, and that commercial-product mentions are not endorsements.

  • The authors’ opinions do not necessarily reflect the views of the U.S. Environmental Protection Agency.
  • The MCDA-MSS database contains 205 methods mapped across 156 characteristics using 31,980 manually assigned binary values.The authors acknowledge that other researchers may disagree with some method mappings and invite corrections.
  • The developers do not take responsibility for choices users make based on MCDA-MSS recommendations.

6. Funding

The paper acknowledges financial support from European Union, Polish National Science Center, Polish Ministry of Science and Higher Education, and related research programmes.

  • The authors acknowledge funding from the European Union’s Horizon 2020 programme, the Polish National Science Center, and Polish Ministry of Science and Higher Education.

Figures of the paper

The paper’s figures and tables illustrate the MCDA-MSS workflow, taxonomy, method matching, recommendation refinement, and detected mismatches across literature case studies.

  • Figure 1 presents the challenge decision analysts face when structuring a decision-making problem.
  • Figures 4–7 show unmatched-method cases, feature selection, recommended methods, missed features, and the most selective questions.
  • The taxonomy covers problem statements, performance comparisons, criteria weights, preference elicitation, aggregation, inconsistency, and preference-model exploitation.
  • Table 2 maps example MCDA methods to taxonomy features and identifies binary values used to trigger decision rules.
  • The case-study tables compare requested features with method capabilities and report complete matches, recommendations, and missed features.
  • The results include recommendations such as DRSA Variable Consistency for sorting, WAM with performances transformation, and GAIA-SMAA-PROMETHEE-INT under specified feature requirements.

Electronic Supplementary Information (ESI)

The supplementary information contains appendices documenting the MCDA-MSS method database, taxonomy mapping, and literature case-study mapping.

  • The ESI lists appendices for the MCDA-MSS method database, taxonomy mapping, and literature case-study mapping.

Appendix A – Methods included in the MCDA-MSS database

Appendix A catalogs MCDA methods included in the MCDA-MSS database, providing brief descriptions and references across diverse decision-problem types and modeling approaches.

  • Table S1 lists MCDA methods in the MCDA-MSS database with brief descriptions and references.
  • Several methods incorporate specialized preference information, including pairwise comparisons, reference-alternative assignments, class cardinalities, and inconsistent assignments.
  • The catalog includes methods for choice, ranking, and sorting problems using scoring, outranking, distance-based, fuzzy-rule, and preference-disaggregation approaches.
  • Recommendation outputs vary across complete or partial rankings, sorting assignments, and cardinal or ordinal scales.
  • The database covers extensions addressing hierarchical or interacting criteria, unknown monotonicity, imprecise performances, and heterogeneous measurement scales.
Loading 2106.07378v1…