Helder Gomes Costa

This study examines whether missing-value imputation in K-means clustering can be reformulated as an optimization problem guided by clustering quality. To address this question, the paper proposes Genetic K-means, a hybrid framework in which a Genetic Algorithm searches the space of candidate imputations and evaluates each solution through the Silhouette score obtained after applying K-means. The method was tested on three benchmark datasets from the UCI repository after inducing artificial missingness under an MCAR mechanism at 5\%, 10\%, and 20\%. Its performance was compared with min-max value replacements, mean filling, and k-Nearest Neighbors imputation using both Silhouette score and RMSE. Across all three datasets and all missingness levels, Genetic K-means achieved the highest Silhouette values, including 0.7154, 0.6744, and 0.5631 for Wholesale Customers, 0.2044, 0.2853, and 0.2433 for Travel Reviews, and 0.2675, 0.2894, and 0.3122 for Cervical Cancer Behavior Risk. In contrast, it did not consistently attain the lowest RMSE. These findings indicate that reconstruction accuracy and clustering adequacy are not equivalent objectives. The originality of the study lies in introducing an imputation framework for unsupervised learning in which evolutionary search is explicitly aligned with structure-preserving clustering performance, thereby extending missing-data treatment beyond conventional reconstruction-oriented logic.

This study addresses limitations of fully compensatory multi-criteria prioritisation by proposing an evidence-weighted, bi-dimensional framework that separates Value (expected public or strategic benefit) from Viability (delivery readiness). By reducing complex MCDA outputs to two governance-relevant dimensions, the approach enhances visual interpretation, supports decisionmakers with predominantly managerial profiles, and makes non-compensatory trade-offs explicit. Globally validated Analytic Hierarchy Process (AHP) weights are used to aggregate raw 1–5 ratings, mapped to a monotone-better scale, into two composite indices (Value and Viability). Proposals are positioned on a governance-oriented 3×3 matrix using tertile thresholds (33rd/67th percentiles) on each axis. The four corner cells (Low×Low, High×Low, Low×High, and High×High) are associated with prescriptive governance actions, cells containing exactly one Medium category are treated as transition bands, and the central Medium×Medium cell is interpreted as a zone of managerial flexibility. New proposals can be incorporated incrementally by scoring and plotting themwithout re-running the entire portfolio. A fully reproducible R workflow computes composite scores and generates the visual outputs. In an illustrative application using a simulated public-healthportfolio (n=39), the bivariate representation reveals Value–Viability imbalances that are masked by single-index rankings. The framework limits compensatory masking, highlights delivery constraints affecting high-value initiatives, and makes explicit when feasible but lower-value proposals may remain attractive under available capacity, while preserving expert-elicited trade-offs within each dimension. Overall, the study bridges managerial visual matrices and MCDA rigor by delivering an auditable, evidence-weighted prioritisation instrument that combines dimensional reduction, visual governance, explicit non-compensatory safeguards, and incremental portfolio extensibility, supported by an open R script implementation.

The SDG Index aggregates country performance across 17 Sustainable Development Goals using arithmetic means, which allows high scores in some goals to compensate for poor performance in others, if weighted sum is applied as the aggregation function. To address this limitation, we develop and apply the KRP2 algorithm that combines K-means clustering with PROMETHEE II ranking to group countries based on their complete SDG performance profiles rather than overall scores alone. Applying this methodology to 2023 data reveals substantial disparities between high-performing and low-performing clusters. Countries in top-ranked categories show strong socioeconomic performance but score lowest on SDG 12 (Responsible Consumption and Production) and SDG 13 (Climate Action). Conversely, bottom-ranked categories struggle with poverty and basic needs (SDGs 1, 7, and 9) yet exhibit the best performance on SDGs 12 and 13, reflecting their minimal historical contribution to environmental degradation. These trade-offs remain hidden when using compensatory aggregation methods. We extend the analysis with a nine-cluster sensitivity test that further isolates extreme cases, confirming the robustness of these patterns. Our framework provides decision-support inputs for policy deliberation, suggesting that differentiated approaches may better serve sustainable development objectives than uniform global policies.

The question that drives this research is: “How to discover the number of respondents that are necessary to validate items of a questionnaire as actually essential to reach the questionnaire’s proposal?” Among the efforts in this subject, Lawshe (Pers. Psychol 28:563–575, 1975), Wilson et al. (Meas. Eval. Couns. Dev. 45:197–210, 2012), Ayre and Scally (Meas. Eval. Couns. Dev. 47:79–86, 2014), Lynn (Nurs. Res. 35:382–386, 1986); Polit et al. (Res. Nurs. Health 30:459–467, 2007) approached this issue by proposing the Content Validation Ratio (CVR) and Content Validations index (CVI) proportion that look to identify items that must be relevant or even essential in a questionnaire. Despite their contribution, these studies do not check if an item validated as “essential” should be also validated as “not essential” by the same sample, which should be a paradox. Another issue is the the loss of nuance caused by assigning a probability equal to 50% to a item be randomly checked by a respondent as essential, despite the rater has three or more options to choose. Our proposal faces these issues, making it possible to verify if a paradoxical situation occurs, and being more precise in recommending whether an item should either be retained or discarded from a questionnaire.

Traditional methods for evaluating airports users’ Level of Satisfaction (LoS) often aggregate customers ratings compensatorily, masking critical weaknesses and ignoring that passengers evaluate services based on different criteria sets. This paper aims to face this issue by proposing a non-compensatory sorting framework that respects passengers’ partial evaluations and identifies improvement issues for airport managers. We inputted the evaluations provided by 19,643 passengers across 15 Brazilian airports into the pessimistic procedure of ELECTRE TRI ME, considering five categories of classification. In a pioneer way, we use the values of the credibility degree to organize airports into groups, resulting in grouping them according to their similarities profile in the credibility degree. This framework provides airport managers with insights to support their decision-making: a transparent LoS classification avoiding compensatory distortions, and cluster-based insights for actions. From a methodological viewpoint, this paper presents an innovative approach to utilizing the complete values calculated for the credibility degree. It also is the first application of ELETRE TRI ME to a huge dataset.

Efficient allocation of resources in public cancer prevention programs requires evidence-based decision-making processes supported by clear and consistent criteria. The objective of this article is to establish an evidence-based model for prioritizing evaluation criteria in cancer prevention projects in Brazil. The model was developed in six stages: (i) comprehensive review of current evaluation practices within Brazilian cancer prevention programs was performed; (ii) multidisciplinary brainstorming sessions identified 19 candidate evaluation criteria; (iii) developed a structured questionnaires to capture expert judgments on each criterion; (iv) panel of national experts evaluated the criteria using a four-point Likert-type scale; (v) Binomial Cut-level method, at a 95% confidence level, was applied to identify statistically validated criteria; (vi) Analytic Hierarchy Process (AHP) was used to assign relative weights through pairwise comparisons, with consistency ratios below 0.10. The results showed that health impact (25.8% of the total weight), policy alignment (17.7%), epidemiological relevance (10.1%), and equity in access (10.1%) together accounted for more than 60% of the attributed importance, while operational aspects such as management plan (0.8%) received lower weight. These percentages represent the relative contribution of each criterion to the final decision, reflecting expert perceptions of their relevance. The integration of BCV and AHP proved effective in structuring prioritization processes in a consistent, replicable, and adaptable way, offering methodological support for rational resource allocation and the formulation of public health policies.

The purpose of this chapter is to identify and compare the main topics addressed by the indicators performance indicators of microfinance institutions (MFIs), which are discussed in literature and used by some of the main rating agencies, in three dimensions of sustainability (financial, social and environmental), according to the Triple Bottom Line perspective. The indicators were identified in the literature (Scopus and Web of Science databases) and in 12 specialized rating agencies. Literature and rating agencies converge on topics like structured financial planning, social impacts, and mitigation of environmental risks. The rating agencies reflect the microfinance market practices with focus on financial factors (such as profitability and productivity) coupled with the MFIs social and environmental objectives. Divergently, literature emphasizes the importance of the accessibility and diversity of microfinance services.