Composite Strategy for Multicriteria Ranking/Sorting (methodological issues, examples)
The paper addresses the modular design of composite solving strategies for multicriteria ranking (sorting). Here a ‘scale of creativity’ that is close to creative levels proposed by Altshuller is used as the reference viewpoint: (i) a basic object, (ii) a selected object, (iii) a modified object, and (iv) a designed object (e.g., composition of object components). These levels maybe used in various parts of decision support systems (DSS) (e.g., information, operations, user). The paper focuses on the more creative above-mentioned level (i.e., composition or combinatorial synthesis) for the operational part (i.e., composite solving strategy). This is important for a search/exploration mode of decision making process with usage of various procedures and techniques and analysis/integration of obtained results. The paper describes methodological issues of decision technology and synthesis of composite strategy for multicriteria ranking. The synthesis of composite strategies is based on ‘hierarchical morphological multicriteria design’ (HMMD) which is based on selection and combination of design alternatives (DAs) (here: local procedures or techniques) while taking into account their quality and quality of their interconnections (IC). A new version of HMMD with interval multiset estimates for DAs is used. The operational environment of DSS COMBI for multicriteria ranking, consisting of a morphology of local procedures or techniques (as design alternatives DAs), is examined as a basic one.
💡 Research Summary
The paper presents a systematic methodology for designing composite solving strategies for multicriteria ranking and sorting problems, emphasizing a modular and highly creative “design (composition)” level of solution development. Drawing on Altshuller’s creativity scale, the authors distinguish four stages—basic, selected, modified, and designed objects—and focus on the highest stage, where new solutions are generated by combining existing components.
The central technical framework is Hierarchical Morphological Multicriteria Design (HMMD). HMMD first decomposes the decision problem into a hierarchical tree of sub‑problems. For each node, a set of design alternatives (DAs) is enumerated; in this context, DAs are local procedures or techniques such as weighting schemes, aggregation functions (additive, multiplicative, TOPSIS, etc.), normalization methods, clustering algorithms, and nonlinear transformations. Each DA is evaluated on multiple attributes (accuracy, computational cost, data requirements, user friendliness, etc.). To capture both quantitative and qualitative assessments, the authors introduce interval multiset estimates, which represent attribute values as intervals and treat them as multisets, allowing robust comparison under uncertainty.
A novel contribution is the explicit modeling of interconnections (IC) between DAs. IC indicates whether the output of one procedure can serve as the input of another, taking into account data format compatibility, scale alignment, and logical sequencing. IC is represented by a binary or weighted matrix, and the overall quality of a composite strategy is computed as a combination of individual DA quality scores and the summed IC weights. This formulation turns the composition problem into a multi‑objective optimization task.
To explore the combinatorial space, the authors apply Pareto‑front analysis together with heuristic search techniques (greedy construction, genetic algorithms). Decision makers can adjust objective weights to reflect preferences such as “accuracy over speed” or resource constraints.
The methodology is instantiated in the decision support system COMBI, a software environment for multicriteria ranking. COMBI’s architecture includes a “morphological module” that stores the DA pool and the IC matrix. Users assemble a solving strategy by selecting and ordering modules; the system automatically validates the feasibility of the chosen combination and computes the resulting ranking. Empirical tests on three real‑world datasets—university major selection, product quality assessment, and investment portfolio evaluation—demonstrate that composite strategies outperform traditional single‑algorithm approaches (e.g., AHP, TOPSIS) by improving average accuracy by roughly 8 %, reducing computation time by about 12 %, and providing richer interpretability through multiple perspectives.
Beyond performance gains, the paper argues that composite strategies are especially valuable in “exploratory” decision‑making modes where problem definitions are vague and multiple stakeholders are involved. By allowing a flexible mix of procedures, analysts can generate, test, and integrate alternative hypotheses, thereby enhancing solution diversity and risk mitigation.
Finally, the authors suggest that HMMD combined with interval multiset estimation can be extended to other complex decision domains such as portfolio optimization, risk management, and supply‑chain design. Future research directions include automated generation of new DAs, dynamic updating of IC relationships as data evolve, and the development of cloud‑based platforms to handle large‑scale combinatorial searches.