Guilherme Dean Pelegrina
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Guilherme Dean Pelegrina has a total of 22 co-authors. Here are the top co-authors by number of shared publications:
- Leonardo Tomazeli Duarte
University of Campinas
19 shared publications João Marcos Travassos Romano
University of Campinas
4 shared publicationsMichel Grabisch
3 shared publications- Michel Grabisch
Centre d’Économie de la Sorbonne
3 shared publications - Miguel Couceiro
INESC-ID, Instituto Superior Técnico, Universidade de Lisboa, Lisbon, Portugal
3 shared publications - Sajid Siraj
Centre for Decision Research Leeds University Business School Leeds UK
3 shared publications
Latest Publications
- Open Access
- 28 August 2026
- Published by Elsevier in SSRN Electronic Journal
A Fairness-Oriented Framework for Parameter Adjustment in Multi-Criteria Decision Analysis: Applications to Linear Aggregation Models and the Choquet Integral
The issues of bias and fairness in decision-making processes within the scope of machine learning have gained increasing attention over the last few years. Motivated by this growing interest, in this paper we address these issues in the context of multi-criteria decision analysis (MCDA). In particular, we provide a framework for guiding the adjustment of the parameters of the aggregation operator in MCDA with the aim of supporting the pursuit of fair rankings, that is, rankings that take into account notions of fairness among groups of alternatives. The proposed framework is developed for aggregation operators given by the simple additive weighting (SAW) method and by the Choquet integral, a non-linear model that has been extensively used in MCDA. At the core of our approach, we rely on the optimization of a fairness metric known as the normalized discounted cumulative Kullback–Leibler divergence (NDKL) with respect to the corresponding parameters of the aggregation operators. In view of the non-convexity of the resulting optimization problem, we adopted the Differential Evolution (DE) meta-heuristic. Numerical experiments were conducted using both synthetic and real-world datasets, demonstrating the effectiveness of the proposed method.
- 28 August 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessA Fairness-Oriented Framework for Parameter Adjustment in Multi-Criteria Decision Analysis: Applications to Linear Aggregation Models and the Choquet Integral
The issues of bias and fairness in decision-making processes within the scope of machine learning have gained increasing attention over the last few years. Motivated by this growing interest, in this paper we address these issues in the context of multi-criteria decision analysis (MCDA). In particular, we provide a framework for guiding the adjustment of the parameters of the aggregation operator in MCDA with the aim of supporting the pursuit of fair rankings, that is, rankings that take into account notions of fairness among groups of alternatives. The proposed framework is developed for aggregation operators given by the simple additive weighting (SAW) method and by the Choquet integral, a non-linear model that has been extensively used in MCDA. At the core of our approach, we rely on the optimization of a fairness metric known as the normalized discounted cumulative Kullback–Leibler divergence (NDKL) with respect to the corresponding parameters of the aggregation operators. In view of the non-convexity of the resulting optimization problem, we adopted the Differential Evolution (DE) meta-heuristic. Numerical experiments were conducted using both synthetic and real-world datasets, demonstrating the effectiveness of the proposed method.
- 18 July 2025
- Published by Springer Nature in Theory and Decision
- Vol. 99 (1-2) , 71-95
Game-theoretic extensions of logistic regression
- 18 July 2025
- Published by Springer Nature in Theory and Decision
- Vol. 99 (1-2) , 71-95
Game-theoretic extensions of logistic regression
- Open Access
- 6 June 2025
- Published by Associacao Brasileira de Estudos Populacionais in Revista Brasileira de Estudos de População
- Vol. 42, 1-30
Spatial distribution of population and forest cover in extractive reserves in the Amazon biome, Brazil
This study relates spatial measures of forest cover with measures of the spatial distribution of the population in the 31 Extractive Reserves (ERs) within the Amazon biome, Brazil, in 2010. We integrated layers of information on the ERs, forest cover, and spatial distribution of the population in a Geographic Information System. We produced 24 variables in three groups: population; external conditions (both as predictor variables); and land cover (predicted variables). We assessed the correlation between predictor variables and forest cover and fragmentation variables. Linear regression analyses based on cooperative game theory were conducted to evaluate the significance of the predictor variables in explaining the number of forest fragments and the percentage of forest cover in the models. We found that the size, concentration, dispersion, and geometry of the population contributed to a better understanding of deforestation and the landscape structure. However, forest fragmentation and forest cover extent are not necessarily defined by the same population aspects. The models suggest that forest cover change is primarily driven by population concentration within the ER, while forest fragmentation is strongly shaped by population dispersion. External conditions such as surrounding forests and protected areas also played a significant role. Our study highlights the importance of incorporating forest spatial distribution measures into Population and Environment research, going beyond the usual focus on forest extent. Additionally, it highlights the value of working with spatial demographic variables, extending beyond the conventional approach centered on population size.
- 6 June 2025
- Published by Associacao Brasileira de Estudos Populacionais in Revista Brasileira de Estudos de População
- Vol. 42, 1-30
Open AccessSpatial distribution of population and forest cover in extractive reserves in the Amazon biome, Brazil
This study relates spatial measures of forest cover with measures of the spatial distribution of the population in the 31 Extractive Reserves (ERs) within the Amazon biome, Brazil, in 2010. We integrated layers of information on the ERs, forest cover, and spatial distribution of the population in a Geographic Information System. We produced 24 variables in three groups: population; external conditions (both as predictor variables); and land cover (predicted variables). We assessed the correlation between predictor variables and forest cover and fragmentation variables. Linear regression analyses based on cooperative game theory were conducted to evaluate the significance of the predictor variables in explaining the number of forest fragments and the percentage of forest cover in the models. We found that the size, concentration, dispersion, and geometry of the population contributed to a better understanding of deforestation and the landscape structure. However, forest fragmentation and forest cover extent are not necessarily defined by the same population aspects. The models suggest that forest cover change is primarily driven by population concentration within the ER, while forest fragmentation is strongly shaped by population dispersion. External conditions such as surrounding forests and protected areas also played a significant role. Our study highlights the importance of incorporating forest spatial distribution measures into Population and Environment research, going beyond the usual focus on forest extent. Additionally, it highlights the value of working with spatial demographic variables, extending beyond the conventional approach centered on population size.
- Open Access
Shapley Value Approximation Based on k-Additive Games
The Shapley value is the prevalent solution for fair division problems in which a payout is to be divided among multiple agents. By adopting a game-theoretic view, the idea of fair division and the Shapley value can also be used in machine learning to quantify the individual contribution of features or data points to the performance of a predictive model. Despite its popularity and axiomatic justification, the Shapley value suffers from a computational complexity that scales exponentially with the number of entities involved, and hence requires approximation methods for its reliable estimation. We propose SVA$k_{\text{ADD}}$, a novel approximation method that fits a $k$-additive surrogate game. By taking advantage of $k$-additivity, we are able to elicit the exact Shapley values of the surrogate game and then use these values as estimates for the original fair division problem. The efficacy of our method is evaluated empirically and compared to competing methods.
Open AccessShapley Value Approximation Based on k-Additive Games
The Shapley value is the prevalent solution for fair division problems in which a payout is to be divided among multiple agents. By adopting a game-theoretic view, the idea of fair division and the Shapley value can also be used in machine learning to quantify the individual contribution of features or data points to the performance of a predictive model. Despite its popularity and axiomatic justification, the Shapley value suffers from a computational complexity that scales exponentially with the number of entities involved, and hence requires approximation methods for its reliable estimation. We propose SVA$k_{\text{ADD}}$, a novel approximation method that fits a $k$-additive surrogate game. By taking advantage of $k$-additivity, we are able to elicit the exact Shapley values of the surrogate game and then use these values as estimates for the original fair division problem. The efficacy of our method is evaluated empirically and compared to competing methods.
- Open Access
Game-theoretic extensions of logistic regression
Logistic regression (LR) is a widely used and well-established tool for classification tasks, valued for its simplicity and interpretability. However, its reliance on linear separators transformed by a sigmoid function limits its flexibility, particularly when dealing with non-linear decision boundaries. To address this limitation, Tehrani et al. proposed an extension of LR where the linear separator is replaced by a piecewise-linear known as Choquet integral. While e!ective, this approach, known as "Choquistic regression", faces the curse of dimensionality, as its parameter count grows exponentially with the number of input features. To mitigate this issue, simplified variants of the Choquet integral, such as the 2-additive Choquet integral, have been proposed. These variants restrict interactions to at most two features, significantly reducing complexity while delivering competitive performance compared to unrestricted models. However, the monotonicity constraints imposed by the Choquistic approach remain a limitation, reducing its applicability to real-world scenarios. In this paper, we propose a novel formulation of logistic regression that integrates the Choquet integral, with parameters defined through a cooperative game framework. Additionally, we explore game-based aggregation functions such as the multilinear model, which also captures feature interactions e!ectively. To overcome the curse of dimensionality, we utilize 2-additive games, which significantly reduce the number of parameters while preserving competitive accuracy. This reduction in complexity is empirically validated through a comparative study of restricted and unrestricted logistic regression variants on a COVID-19 dataset. Furthermore, the results highlight the interpretability advantages of these models, demonstrating their practical value in real-world applications.
Open AccessGame-theoretic extensions of logistic regression
Logistic regression (LR) is a widely used and well-established tool for classification tasks, valued for its simplicity and interpretability. However, its reliance on linear separators transformed by a sigmoid function limits its flexibility, particularly when dealing with non-linear decision boundaries. To address this limitation, Tehrani et al. proposed an extension of LR where the linear separator is replaced by a piecewise-linear known as Choquet integral. While e!ective, this approach, known as "Choquistic regression", faces the curse of dimensionality, as its parameter count grows exponentially with the number of input features. To mitigate this issue, simplified variants of the Choquet integral, such as the 2-additive Choquet integral, have been proposed. These variants restrict interactions to at most two features, significantly reducing complexity while delivering competitive performance compared to unrestricted models. However, the monotonicity constraints imposed by the Choquistic approach remain a limitation, reducing its applicability to real-world scenarios. In this paper, we propose a novel formulation of logistic regression that integrates the Choquet integral, with parameters defined through a cooperative game framework. Additionally, we explore game-based aggregation functions such as the multilinear model, which also captures feature interactions e!ectively. To overcome the curse of dimensionality, we utilize 2-additive games, which significantly reduce the number of parameters while preserving competitive accuracy. This reduction in complexity is empirically validated through a comparative study of restricted and unrestricted logistic regression variants on a COVID-19 dataset. Furthermore, the results highlight the interpretability advantages of these models, demonstrating their practical value in real-world applications.
- 9 October 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Fuzzy Systems
- Vol. 32 (12) , 7196-7201
A Novel Unsupervised Capacity Identification Approach to Deal With Redundant Criteria in Multicriteria Decision Making Problems
The use of the Choquet integral in multicriteria decision making problems has gained attention in the last two decades. Despite of its usefulness, there is the issue of how to define the Choquet integral parameters, called capacity coefficients, specially the ones associated with coalitions of criteria. A possible approach to address this issue is based on unsupervised learning, which aims to define such parameters with the goal of mitigating undesirable effects provided by intercriteria relations. However, current unsupervised approaches present some drawbacks, as there is no guarantee that the parameters are equally prioritized in the learning procedure. In this paper, we propose a novel unsupervised capacity identification approach which ensures a fair learning for all parameters. Moreover, in comparison with the existing methods, our proposal is less complex in terms of optimization, as it is based on a linear formulation. Experimental results in both synthetic and real datasets attest the applicability of our proposal.
- 9 October 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Fuzzy Systems
- Vol. 32 (12) , 7196-7201
A Novel Unsupervised Capacity Identification Approach to Deal With Redundant Criteria in Multicriteria Decision Making Problems
The use of the Choquet integral in multicriteria decision making problems has gained attention in the last two decades. Despite of its usefulness, there is the issue of how to define the Choquet integral parameters, called capacity coefficients, specially the ones associated with coalitions of criteria. A possible approach to address this issue is based on unsupervised learning, which aims to define such parameters with the goal of mitigating undesirable effects provided by intercriteria relations. However, current unsupervised approaches present some drawbacks, as there is no guarantee that the parameters are equally prioritized in the learning procedure. In this paper, we propose a novel unsupervised capacity identification approach which ensures a fair learning for all parameters. Moreover, in comparison with the existing methods, our proposal is less complex in terms of optimization, as it is based on a linear formulation. Experimental results in both synthetic and real datasets attest the applicability of our proposal.
- 17 July 2024
- Published by Elsevier in Expert Systems with Applications
- Vol. 255
Mitigating subjectivity and bias in AI development indices: A robust approach to redefining country rankings
- 17 July 2024
- Published by Elsevier in Expert Systems with Applications
- Vol. 255
Mitigating subjectivity and bias in AI development indices: A robust approach to redefining country rankings
- Open Access
- 3 June 2024
- Published by Association for Computing Machinery (ACM)
- p. 279-289
A preprocessing Shapley value-based approach to detect relevant and disparity prone features in machine learning
Decision support systems became ubiquitous in every aspect of human lives. Their reliance on increasingly complex and opaque machine learning models raises transparency and fairness concerns with respect to unprivileged groups of people. This motivated several efforts to estimate importance of features towards the models’ performance and to detect unfair/disparate decisions. The latter is often dealt with by means of fairness metrics that rely on performance metrics with respect to predefined features that are considered protected (salient features such as age, gender, ethnicity, etc.) and/or sensitive (such as education, /occupation, banking information). However, such an approach is subjective (as fairness metrics depend on the choice features), there may be other features that lead to unfair (disparate) decisions and that may ask for suitable interpretations. In this paper we focus on the latter issues and propose a statistical preprocessing approach that is inspired by both the Hilbert-Schmidt independence criterion and Shapley values to estimate feature importance and to detect disparity prone features. Unlike traditional Shapley value-based approaches, we do not require trained models to measure feature importance or detect disparate results. Instead, it focuses on data and statistical criteria to measure the dependence of feature distributions. Our empirical results show that features with the highest dependence degrees with the label vector are also the ones with the highest impact on the model performance. Moreover, our empirical results indicate that this relation enables the detection of disparity prone features.
- 3 June 2024
- Published by Association for Computing Machinery (ACM)
- p. 279-289
Open AccessA preprocessing Shapley value-based approach to detect relevant and disparity prone features in machine learning
Decision support systems became ubiquitous in every aspect of human lives. Their reliance on increasingly complex and opaque machine learning models raises transparency and fairness concerns with respect to unprivileged groups of people. This motivated several efforts to estimate importance of features towards the models’ performance and to detect unfair/disparate decisions. The latter is often dealt with by means of fairness metrics that rely on performance metrics with respect to predefined features that are considered protected (salient features such as age, gender, ethnicity, etc.) and/or sensitive (such as education, /occupation, banking information). However, such an approach is subjective (as fairness metrics depend on the choice features), there may be other features that lead to unfair (disparate) decisions and that may ask for suitable interpretations. In this paper we focus on the latter issues and propose a statistical preprocessing approach that is inspired by both the Hilbert-Schmidt independence criterion and Shapley values to estimate feature importance and to detect disparity prone features. Unlike traditional Shapley value-based approaches, we do not require trained models to measure feature importance or detect disparate results. Instead, it focuses on data and statistical criteria to measure the dependence of feature distributions. Our empirical results show that features with the highest dependence degrees with the label vector are also the ones with the highest impact on the model performance. Moreover, our empirical results indicate that this relation enables the detection of disparity prone features.
- 21 February 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Artificial Intelligence
- Vol. 5 (8) , 4217-4231
Shapley Value-Based Approaches to Explain the Quality of Predictions by Classifiers
The use of algorithm-agnostic approaches for explainable machine learning is an emerging area of research. When explaining the contribution of features towards the predicted outcome, traditionally, the focus remains on explaining the prediction itself, however a little has been done on explaining the quality of prediction of these models, where the quality can be assessed by the algorithm performance when changing the thresholds for classification. In this paper, we propose the use of Shapley values to explain the contribution of features towards the overall algorithm performance, measured in terms of Receiver-operating Characteristics (ROC) curve and the Area under the ROC curve (AUC). With the help of an illustrative example, we demonstrate the proposed idea of explaining the ROC curve, and visualising the uncertainties in these curves. For imbalanced datasets, the use of Precision-Recall Curve (PRC) is considered more appropriate, therefore we also demonstrate how to explain the PRCs with the help of Shapley values. The explanation of the model performance can help analysts in a number of ways, for example, in feature selection by identifying the irrelevant features that can be removed to reduce the computational complexity. It can also help in identifying the features having critical contributions towards the overall algorithm performance.
- 21 February 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Artificial Intelligence
- Vol. 5 (8) , 4217-4231
Shapley Value-Based Approaches to Explain the Quality of Predictions by Classifiers
The use of algorithm-agnostic approaches for explainable machine learning is an emerging area of research. When explaining the contribution of features towards the predicted outcome, traditionally, the focus remains on explaining the prediction itself, however a little has been done on explaining the quality of prediction of these models, where the quality can be assessed by the algorithm performance when changing the thresholds for classification. In this paper, we propose the use of Shapley values to explain the contribution of features towards the overall algorithm performance, measured in terms of Receiver-operating Characteristics (ROC) curve and the Area under the ROC curve (AUC). With the help of an illustrative example, we demonstrate the proposed idea of explaining the ROC curve, and visualising the uncertainties in these curves. For imbalanced datasets, the use of Precision-Recall Curve (PRC) is considered more appropriate, therefore we also demonstrate how to explain the PRCs with the help of Shapley values. The explanation of the model performance can help analysts in a number of ways, for example, in feature selection by identifying the irrelevant features that can be removed to reduce the computational complexity. It can also help in identifying the features having critical contributions towards the overall algorithm performance.