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.

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.

This study addresses discontinuity prediction in income reporting within the Chilean student loan program, a critical event for credit risk management. Although the literature has incorporated machine learning models to anticipate non-compliance behavior, a gap remains in the development of methodologically robust evaluations that integrate nonlinear imputation, imbalance correction, and repeated validation across multiple partitions. To address this need, a complete pipeline was implemented on a dataset of 22,303 records, including MissForest imputation, SMOTE-based balancing, and a comparative assessment of a biologically inspired Deep Neural Network (DNN) and a Random Forest (RF) classifier used as a classical baseline model, evaluated across 35 stratified partitions. The results show that the bioinspired DNN, as the primary focus of this study, consistently outperforms the RF in metrics such as AUC (0.9991 vs 0.9709), F1-score (0.9966 vs 0.9497), and agreement measures, while also exhibiting lower variability across partitions. The interpretability analysis indicates that financial variables account for the greatest influence on predictions, whereas demographic variables contribute minimally. The study provides a replicable and robust methodology aligned with risk analysis practices in student credit contexts.

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.

This work presents a novel dataset designed to support predictive modeling of income declaration behavior in the University Credit Solidarity Fund (FSCU) program of Chile. Using institutional records from the Pontificia Universidad Católica de Valparaíso, we prepared a structured and anonymized dataset that captures academic, financial, and demographic variables relevant to student loan administration. We describe the data extraction and preprocessing pipeline, including entity modeling, feature engineering, and the creation of a relational database. As a baseline evaluation, we conducted initial experiments using 15 machine learning algorithms via PyCaret, aiming to predict whether a student who previously submitted income declarations would cease to do so. Although no hyperparameter tuning or data balancing was applied, the results offer a solid foundation for future experimentation. Boosting-based classifiers-particularly XGBoost-achieved the best performance, with accuracy and AUC exceeding 96 %. This contribution addresses the lack of publicly available, context-sensitive datasets for modeling conditional repayment schemes in higher education finance. Future work includes dataset publication, class balancing, and explainability analysis.