Solving problems associated with the efficient distribution and organization of resources has generated increasing interest in the scientific community. One of the most commonly used approaches consists of approximate solution techniques, which have been able to solve complex covering problems within acceptable computational time and cost. One of the benchmarks used to evaluate these approaches is the Set Covering Problem, which is an NP-hard combinatorial optimization problem. Among the techniques that have been investigated, metaheuristics play an important role. These methods are commonly developed for continuous search spaces and, in order to be applied to covering problems, must be modified to operate in discrete domains. This modification presents an important challenge: finding an appropriate transformation method that translates continuous solutions into binary solutions. This issue has been addressed through two main strategies: binarization using two-step schemes, and, in our proposal, the use of repair operators orchestrated according to their performance through an Adaptive Repair Selection Mechanism based on the multi-armed bandit framework. To evaluate our proposal, we selected the Binary Hunger Games Search metaheuristic because the relative quality of each individual determines its hunger level, which in turn regulates the movement of the population and the influence of the best solution found. Infeasible solutions are handled through a set of Tabu Search-based repair operators. Instead of applying a single repair rule throughout the entire execution, the proposed approach dynamically selects among these operators according to their observed contribution during the search. Each repair operator also incorporates Tabu memory to discourage repetitive decisions during feasibility restoration. The experiments were conducted using the classical Beasley benchmark instances for the Set Covering Problem.

This study addresses an interpretable supervised binary classification problem under constrained feature availability and class imbalance. The objective is to evaluate whether reliable predictive performance can be achieved using exclusively pre-event administrative variables while preserving transparency and analytical traceability of model decisions. A comparative framework is developed using linear and ensemble-based classifiers, combined with resampling strategies and exhaustive hyperparameter optimization embedded within cross-validation. Model performance is evaluated using standard classification metrics, with particular emphasis on the Matthews correlation coefficient as a robust measure under imbalance. In addition to predictive accuracy, the analysis incorporates global, structural, and local interpretability mechanisms, including permutation feature importance, explicit decision paths derived from tree-based models, and additive local explanations. Experimental results show that optimized ensemble models achieve consistent performance gains over linear baselines while maintaining a balanced error structure across classes. Importantly, the most influential predictors exhibit stable rankings across models and explanation methods, indicating a concentrated and robust discriminative signal within the constrained feature space. The interpretability analysis demonstrates that complex classifiers can be decomposed into verifiable decision rules and locally coherent feature contributions. Overall, the findings confirm that interpretable supervised classification can be reliably conducted under administrative data constraints, providing a reproducible modeling framework that balances predictive performance, error analysis, and explainability in applied mathematical settings.

The Set Covering Problem is a fundamental NP-hard problem in combinatorial optimization and plays a central role in a wide range of industrial decision-making processes, including logistics planning, scheduling, facility location, network design, and resource allocation. In many real-world contexts, problems of this type are large in scale and highly constrained, which makes exact solution methods computationally impractical and encourages the use of metaheuristic approaches capable of producing high-quality solutions within limited time budgets. In this work, we propose a discrete adaptation of the Dream Optimization Algorithm, focusing on the challenges that emerge when algorithms originally designed for continuous search spaces are applied to binary and strongly constrained models. The continuous search process is mapped onto the binary decision space through a fixed discretization scheme. As a consequence of this transformation, some constraints may not be met, underscoring the importance of effective feasibility restoration mechanisms. Because the discretization stage may produce infeasible solutions and frequently induces plateaus that hinder further improvement, an explicit repair phase becomes necessary to restore feasibility and promote effective search progression. To strengthen this process, the study introduces an adaptive control mechanism based on bandit driven operator selection, which dynamically chooses among different repair procedures during the search. Experimental results on benchmark instances show that the proposed approach consistently achieves high quality solutions with low relative deviation from known optima and stable behavior across independent runs.

Bio-inspired swarm metaheuristic algorithms constitute a widely used tool for solving complex optimization problems. However, the experimental characterization of their emergent behavior remains a methodological challenge. This study proposes an experimental framework for characterizing emergent behavior through swarm collective dynamics. The framework integrates complementary dynamic indicators and establishes relative diversity loss as a homogeneous criterion for defining equivalent comparison states across different search processes. The framework was evaluated using the Reptile Search Algorithm (RSA) and Draco Lizard Optimizer (DLO) as case studies, with Particle Swarm Optimization (PSO) serving as a reference algorithm. The results showed that swarm collective dynamics were associated with both the mathematical properties of the search landscape and the search mechanisms of each metaheuristic. Furthermore, relative diversity loss enabled the comparison of different metaheuristics within a common reference framework. In RSA, swarm reorganization occurred during the first iterations. DLO exhibited a more gradual evolution, whereas PSO showed an intermediate behavior between both dynamics. The proposed experimental framework provides a methodological basis for the experimental characterization of emergent behavior in swarm metaheuristics.

The non-linearity of symmetric block cryptographic algorithms, which is crucial for resisting linear and differential cryptanalysis, depends on the design of substitution boxes. This work proposes a novel scheme, named GOM, combining opposition-based learning with graph-based representation to be integrated into three population-based metaheuristics. This scheme is applied to address the 8 x 8 S-box design problem. The GOM-enhanced metaheuristics improve population diversity and convergence, achieving a non-linearity of 112. Hardware simulations indicate that GOM-based designs require similar resources to AES, while side-channel evaluations confirm their resilience against power analysis attacks. Scalability is supported by successful results on 10 x 10 and 12 x 12 S-boxes. The design of GOM, leveraging generalizable components such as opposition-based learning and graph-based representations, suggests its potential applicability to other population-based metaheuristics and optimization problems.

The Knapsack Problem (KP) is a classic combinatorial optimization challenge with applications in several fields including logistics and finance. Traditional solving approaches often struggle to solve large-scale and complex instances, necessitating the adoption of advanced metaheuristic techniques. This study presents a novel approach that integrates chaotic maps into the Tasmanian Devil Optimization (TDO) algorithm to address the KP more effectively. Specifically, a chaotic binarization mechanism is proposed to improve the balance between exploration and exploitation. Comparative analyses with state-of-the-art metaheuristics were performed, using KP benchmark datasets. The results demonstrate that the chaotic sine-based TDO outperforms traditional methods in terms of solution quality, convergence stability, and computational efficiency. These findings highlight the potential of chaotic hybridization as a promising tool for solving binary combinatorial optimization problems.

Bio-inspired metaheuristic optimization offers flexible search mechanisms for high-dimensional predictive problems under operational constraints. In administrative risk prediction settings, class imbalance and feature redundancy challenge conventional learning pipelines. This study evaluates a wrapper-based metaheuristic feature selection framework for post-compliance income declaration prediction using real longitudinal administrative records. The proposed approach integrates swarm-inspired optimization with supervised classifiers under a weighted objective function jointly prioritizing minority-class recall and subset compactness. Robustness is assessed through 31 independent stochastic runs per configuration. The empirical results indicate that performance effects are learner-dependent. For variance-prone classifiers, substantial minority-class recall gains are observed, with recall increasing from 0.284 to 0.849 for k-nearest neighbors and from 0.471 to 0.932 for Random Forest under optimized configurations. For LightGBM, optimized models maintain high recall levels (0.935-0.943 on average) with low dispersion, suggesting representational stabilization and dimensional compression rather than large absolute recall improvements. Optimized subsets retain approximately 16-33 features on average from the original 76-variable space. Within the evaluated experimental protocol, the findings show that metaheuristic-driven wrapper feature selection can reshape predictive representations under class imbalance, enabling simultaneous control of minority-class performance and feature dimensionality. Formal institutional deployment and cross-domain generalization remain subjects for future investigation.

Timely graduation, time-to-degree, and degree completion are key indicators of student progression and institutional effectiveness in higher education. This study presents a PRISMA-based systematic literature review of machine learning approaches for graduation-related prediction, with attention to predictive targets, pipeline components, scalability, and bio-inspired optimization. Searches in Web of Science Core Collection and Scopus identified 278 records, of which 25 studies published between 2021 and 2025 met the eligibility criteria. The findings show that most studies formulated graduation prediction as a supervised classification task, relied heavily on academic performance variables, and frequently used tree-based or ensemble models. Feature selection, explainability, and hyperparameter optimization were commonly reported, but bio-inspired optimization was actively implemented in only two studies through Particle Swarm Optimization, Genetic Algorithms, or Ant Colony Optimization. The evidence base also remains limited in scalability, as most studies used single-institution datasets and provided little external validation. These findings identify an opportunity for Bio-Inspired Educational Analytics through scalable feature selection, efficient hyperparameter optimization, model simplification, and multi-objective trade-off analysis. Future research should evaluate whether lightweight, hybrid, and multi-objective metaheuristics can support accurate, interpretable, fair, and transferable graduation prediction systems.

In recent decades, various techniques have been developed to support structural health monitoring (SHM). These approaches have benefited significantly from advances in artificial intelligence and big data, which have enabled the detection of structural damage with increasing accuracy. Although such methods allow for automated and real-time assessment of structural integrity, their deployment is often constrained by hardware requirements because of the high computational demands of machine learning (ML) models. These limitations may hinder the practical implementation of automated SHM systems in real-world scenarios. Although pruning techniques aimed at reducing model size by removing specific components of the neural network have been proposed to address this issue, they often suffer from limitations that may be mitigated through alternative approaches, which remain underexplored. In this study, we propose a novel pruning method that integrates metaheuristic optimization techniques to identify the optimal set of neurons to prune. The proposed approach is evaluated using a real-world case study in which the size of an ML model trained to detect structural damage is reduced. The results show that the proposed technique can reduce the model size by 45%, with only a 0.16 and 0.43% decrease in damage detection performance, measured in terms of F1 score and Area Under Curve (AUC) score, respectively.

Melanoma is one of the most aggressive forms of skin cancer, with a high mortality rate when not detected early. This public health challenge underscores the need for accurate and efficient diagnostic tools. Convolutional Neural Networks have shown strong performance in medical image analysis. However, their effectiveness relies heavily on optimal architectural and hyperparameter configurations, which are often designed without alignment to the target domain or transferred from unrelated domains, limiting adaptability to specific medical datasets. Existing hybrid CNN-metaheuristic approaches typically optimize only fixed network parameters. They often fail to explore how metaheuristics can adaptively shape the CNN architectures themselves.In this study, a comprehensive hybrid optimization framework is proposed that integrates CNNs with six nature-inspired metaheuristic algorithms that mimic biological or physical phenomena to solve complex problems. These include Cuckoo Search, Firefly Algorithm, Whale Optimization Algorithm, Particle Swarm Optimization, Grey Wolf Optimizer, and Crow Search Algorithm. Rather than tuning a predefined architecture, each optimizer searches the architectural and training space to identify high-performing CNN configurations, enabling emergent and data-driven network design. This unified framework allows a systematic cross-algorithm comparison under identical conditions, providing new insights into convergence stability, exploration-exploitation dynamics, and generalization behavior. A robust preprocessing and data augmentation pipeline, including brightness normalization, hair artifact removal, and geometric transformations, is incorporated to improve model generalization and enhance the optimizer's search landscape. Experiments on the HAM10000 dataset demonstrate that the metaheuristic-optimized CNNs outperform the baseline, achieving accuracies up to 91.25%. These findings confirm that population-based optimization is an efficient and reliable mechanism for guiding CNN architecture design. This approach achieves superior performance compared to traditional manual or other optimization-based strategies.

The transition from continuous swarm intelligence algorithms to discrete combinatorial domains remains a critical challenge in bio-inspired computing. Traditional binarization techniques frequently induce premature convergence in highly constrained landscapes. This paper presents a chaotic discretization framework that replaces the classical behavior of the two-step binarization technique to regulate the balance between exploration and exploitation. The proposal systematically integrates three leading continuous metaheuristics in the literature, with twenty-four binarization configurations, across three distinct NP-hard problem archetypes: capacity-constrained (0-1 Knapsack), sparse (Set Covering), and mathematically degenerate flat landscapes (Unicost Set Covering). Nonparametric statistical tests confirm that chaotic discretization acts as a powerful regulator in the landscape (p < 0.05). Empirical evidence shows that the highest-performing chaotic mapping is heavily influenced by the specific landscape morphology evaluated: the 0-1 Knapsack Problem is statistically optimized by the Circle map under standard rules; the Set Covering Problem achieves optimal median performance with the Tent map under elitist formulations, although severe matrix constraints ultimately force statistical ties; and the Unicost Set Covering Problem utilizes the nonlinear sequences of the sinusoidal map under complementary operators to break convergence stagnation.

Background: Viral Pneumonia and Tuberculosis continue to represent a significant burden on global public health, relying heavily on chest X-rays for screening and diagnosis. Although deep learning systems offer promising diagnostic support, the traditional manual tuning of hyperparameters for Convolutional Neural Networks is often inefficient and computationally expensive, frequently resulting in suboptimal or overly heavy architectures. Methods: To address these challenges, this study proposes a hybrid framework that employs metaheuristic algorithms, specifically the Whale Optimization Algorithm, Grey Wolf Optimizer, and Cuckoo Search to automatically optimize the architecture and training parameters of a custom neural network for the multi-class classification of Normal, Viral Pneumonia, and Tuberculosis cases. The proposed approach was evaluated using a rigorous stratified k-fold cross-validation protocol on a balanced, multi-source dataset. Results: The experimental results demonstrate that the model optimized by the Whale Optimization Algorithm statistically outperforms manually configured baselines, achieving the highest diagnostic accuracy and specificity. Furthermore, a critical finding of this research is the substantial improvement in computational efficiency; the automated optimization reduced the computational load by approximately 74% and the storage requirements by 63%, making the model viable for deployment in resource-constrained environments. Conclusions: Finally, to ensure clinical reliability, the decision-making process was validated using Gradient-weighted Class Activation Mapping, which confirmed that the network successfully learns to identify clinically relevant pulmonary structures while ignoring confounding artifacts.

Online parameter tuning significantly enhances the performance of optimization algorithms by dynamically adjusting mutation and crossover rates. However, current approaches often suffer from high computational costs and limited adaptability to complex and dynamic fitness landscapes, particularly when machine learning methods are employed. This work proposes a quantized shallow neural network (SNN) as an efficient learning-based component for dynamically adjusting the mutation and crossover rates of a genetic algorithm (GA). By leveraging runtime-generated data and applying quantization techniques like Quantization-aware Training (QaT) and Post-training Quantization (PtQ), the proposed approach reduces computational overhead while maintaining competitive performance. Experimental evaluation on 15 continuous benchmark functions demonstrates that the quantized SNN achieves high-quality solutions while significantly reducing execution time compared to alternative shallow learning methods. This study highlights the potential of quantized SNNs to balance efficiency and performance, broadening the applicability of shallow learning in optimization.

The Unicost Set Covering Problem (USCP), an NP-hard combinatorial optimization challenge, demands efficient methods to minimize the number of sets covering a universe. This study introduces a binary White Shark Optimizer (WSO) enhanced with V3 transfer functions, elitist binarization, and chaotic maps. To evaluate algorithm performance, we employ the Relative Percentage Deviation (RPD), which measures the percentage difference between the obtained solutions and optimal values. Our approach achieves outstanding results on six benchmark instances: WSO-ELIT_CIRCLE delivers an RPD of 0.7% for structured instances, while WSO-ELIT_SINU attains an RPD of 0.96% in cyclic instances, showing empirical improvements over standard methods. Experimental results demonstrate that circle chaotic maps excel in structured problems, while sinusoidal maps perform optimally in cyclic instances, with observed improvements up to 7.31% over baseline approaches. Diversity and convergence analyses show structured instances favor exploitation-driven strategies, whereas cyclic instances benefit from adaptive exploration. This work establishes WSO as a robust metaheuristic for USCP, with applications in resource allocation and network design.

This paper comprehensively analyzes objective functions used in feature selection, a critical aspect of machine learning. We conducted a systematic literature review, categorizing objective functions into single-objective and multi-objective, with further classification into pure and weighted multi-objective functions. Our study spans research from 2019 to 2023, analyzing 161 articles. We found that weighted multi-objective functions are most prevalent, highlighting their efficacy in balancing model performance and complexity. This work offers a detailed classification of these functions, contributing to a deeper understanding of their role and effectiveness in feature selection challenges. Our findings illuminate trends and preferences in objective function usage, providing valuable insights for researchers and practitioners in machine learning.

Preventive maintenance efforts for bridge infrastructure have proven to mitigate early deterioration and reduce the probability of severe damage. Modern research has focused on the employment of online data directly collected within the structures, provided by several novel devices that feed machine learning approaches that continuously measure structural health. However, several issues remain within the related fields. The constant evolution of ML techniques, for example, provides new potential lines of research. Furthermore, widespread validation through real-world test cases and the use of diverse bridge typologies (which can be interesting considering their distinct behaviors) remain limited. This article seeks to examine the advancements in structural health monitoring (SHM) employing machine learning methods for identifying structural damage in bridges over a 7-year period, with a particular focus on studies employing real bridge data. Present challenges and future research directions are assessed. This study offers valuable insights to researchers and academics conducting research in this field, providing a thorough summary of current developments and combining ML methods with the four most-investigated bridge types in case studies.

Binarizing continuous metaheuristics to solve challenging NP-hard binary optimization problems is a fundamental step in adapting continuous algorithms for discrete domains. Binary optimization problems, such as the Set Covering Problem and the 0–1 Knapsack Problem, demand tailored approaches to efficiently explore and exploit the solution space. The process of binarization often introduces complexities, as it requires balancing the transformation of continuous populations into binary solutions while preserving the algorithm’s capability to navigate the search space effectively. In this context, we explore the performance of the Reptile Search Algorithm (RSA), a continuous metaheuristic, applied to these two benchmark problems. To address the binary nature of the problems, a two-step binarization process is implemented, utilizing combinations of transfer functions with binarization rules. This framework enables the RSA to generate binary solutions while leveraging its inherent strengths in exploration and exploitation. Comparative experiments are conducted with Particle Swarm Optimization and the Grey Wolf Optimizer to benchmark the RSA’s performance under similar conditions. These experiments analyze critical factors such as fitness values, convergence behavior, and exploration–exploitation dynamics, providing insights into the effectiveness of different binarization approaches. The results demonstrate that the RSA achieves competitive performance across both problems, highlighting its flexibility and adaptability, which are attributed to its diverse movement equations. Notably, the Z4 transfer function consistently enhances performance for all algorithms, even when paired with less effective binarization rules. This indicates the potential of Z4 as a robust transfer function for binary optimization. The findings underscore the importance of selecting appropriate binarization strategies to maximize the performance of continuous metaheuristics in binary domains, paving the way for further advancements in hybrid optimization methodologies.

The balance between exploration and exploitation is a fundamental element in the design and performance of bio-inspired optimization algorithms. However, to date, its conceptual evolution and its treatment in the scientific literature have not been systematically characterized from a bibliometric approach. This study performs an exhaustive analysis of the scientific production on the balance between exploration and exploitation using records extracted from the Web of Science (WoS) database. The processing and analysis of the data were carried out through the combined use of Bibliometrix (R package) and VOSviewer, tools that made it possible to quantify productivity, map collaborative networks, and visualize emerging thematic trends. The results show a sustained growth in the volume of publications over the last decade, as well as the consolidation of academic collaboration networks and the emergence of new thematic lines in the field. In particular, metaheuristic algorithms have demonstrated a significant and growing impact, constituting a fundamental pillar in the advancement and methodological diversification of the exploration–exploitation balance. This work provides a quantitative framework and a structured view of the evolution of research, identifies the main actors and trends, and raises opportunities for future lines of research in the field of optimization using metaheuristics, the most prominent instantiation of bio-inspired optimization algorithms.

The Set Coverage Problem (SCP) is an important combinatorial optimization problem known to be NP-complete. The use of metaheuristics to solve the SCP includes different algorithms. In particular, binarization techniques have been explored to adapt metaheuristics designed for continuous optimization problems to the binary domain of the SCP. In this work, we present a new approach to solve the SCP based on the Secretary Bird Optimization Algorithm (SBOA). This algorithm is inspired by the natural behavior of the secretary bird, known for its ability to hunt prey and evade predators in its environment. Since the SBOA was originally designed for optimization problems in continuous space and the SCP is a binary problem, this paper proposes the implementation of several binarization techniques to adapt the algorithm to the discrete domain. These techniques include eight transfer functions and five different discretization methods. Taken together, these combinations create multiple SBOA adaptations that effectively balance exploration and exploitation, promoting an adequate distribution in the search space. Experimental results applied to the SCP together with its variant Unicost SCP and compared to Grey Wolf Optimizer and Particle Swarm Optimization suggest that the binary version of SBOA is a robust algorithm capable of producing high quality solutions with low computational cost. Given the promising results obtained, it is proposed as future work to focus on complex and large-scale problems as well as to optimize their performance in terms of time and accuracy.

Metaheuristics have proven to be effective in solving large-scale combinatorial problems by combining global exploration with local exploitation, all within a reasonably short time. The balance between these phases is crucial to avoid slow or premature convergence. We propose binary variants of the Arithmetic Optimization Algorithm for the set cover problem, integrating a two-step binarization scheme based on transfer functions with binarization rules and a greedy repair operator to ensure feasibility. We evaluate the proposed solution using forty-five instances from OR-Beasley and compare it with representative approaches, including genetic algorithms, path-relinking strategies, and Lagrangian-based heuristics. The quality of the solution is evaluated using relative percentage deviation and stability with the coefficient of variation. The results show competitive deviations and consistently low variation, confirming that our approach is a robust alternative with a solid balance between exploration and exploitation.