Guillermo Sanchez-Diaz

Constructs in rough set theory extend attribute reduction beyond inter-class discernibility by also preserving similarity among objects associated with the same decision class. Although class-specific reducts have been systematically studied, an analogous one-versus-the-rest formulation for constructs requires additional analysis because merging the non-target classes introduces new similarity-preservation requirements. This paper introduces class-specific attribute constructs within a preservation-based framework in which the complete set of descriptive attributes is taken as the reference. For each target class, the original multiclass problem is transformed into a binary target-versus-complement decision problem, and a class-specific construct is defined as an inclusion-minimal subset preserving the corresponding discernibility and similarity requirements. The relationships between conventional and class-specific constructs are characterized through two sets of object pairs, which describe the similarity requirements introduced and the discernibility requirements removed by the one-versus-the-rest transformation. Conditions for containment and exact union representation are established, together with aggregation and family-level results that clarify which properties of class-specific reducts extend to constructs and which do not. Experiments on the Seeds and Wine datasets illustrate that class-specific constructs may be shorter, comparable, or longer than conventional constructs, producing corresponding changes in rule coverage, selectivity, and class-dependent classification behavior.

Feature selection methods in Rough Set Theory and Logical Combinatorial Pattern Recognition commonly rely on simplified binary discernibility matrices to represent the condition attributes that distinguish objects from different decision classes. However, the original discernibility matrix involves a large number of redundant object comparisons, which severely limits computational efficiency and scalability. Although several algorithms have been proposed to compute the simplified matrix, most of them first construct the full discernibility matrix and then simplify it, making these approaches impractical for large-scale datasets. In this paper, we introduce ISBMA, an incremental algorithm based on a deterministic finite automaton that directly constructs the simplified binary discernibility and discernibility/similarity matrices without explicitly generating the full matrix. Experimental comparisons with the fastest reported method demonstrate that ISBMA significantly reduces redundant computations, requires substantially less memory, and achieves superior runtime performance, remaining feasible on large decision tables where existing approaches fail.

Corrosion of steel reinforcement is the major factor that limits the durability and serviceability performance of reinforced concrete structures. Impressed current cathodic protection (ICCP) is a widely used method to protect steel reinforcements against corrosion. This research aimed to study the effect of cathodic protection on reinforced concrete with fly ash using electrochemical noise (EN). Two types of reinforced concrete mixtures were manufactured; 100% Ordinary Portland Cement (OCP) and replacing 15% of cement using fly ash (OCPFA). The specimens were under-designed protected conditions (−1000 ≤ E ≤ −850 mV vs. Ag/AgCl) and cathodic overprotection (E < −1000 mV vs. Ag/AgCl) by impressed current, and specimens concrete were immersed in a 3.5 wt.% sodium chloride (NaCl) Solution. The analysis of electrochemical noise-time series showed that the mixtures microstructure influenced the corrosion process. Transients of uniform corrosion were observed in the specimens elaborated with (OPC), unlike those elaborated with (OPCFA). This phenomenon marked the difference in the concrete matrix’s hydration products, preventing Cl⁻ ions flow and showing passive current and potential transients in most specimens.

Dengue is a major public health concern mainly in tropical and subtropical environments worldwide. Despite several attempts to prevent this disease occurring in tropical regions of Mexico, it has not yet been controlled. This work focused on spatial modeling of confirmed dengue fever cases that occurred during the period 2010–2014 in the Huasteca Potosina region of Mexico. Multivariable Logistic Regression Modeling (MLRM) was used to determine the relationship between explanatory variables and the presence/absence of dengue. Model performance was evaluated using the area under curve (AUC) of the relative operating characteristic (ROC); AUC > 0.95. A high spatial resolution map was created to reveal the most probable patterns of dengue risk. Our results can be used for targeted control and prevention programs at local and regional levels. This methodology can be applied to other major diseases that are spatially distributed in accordance with environmental factors.

La zona metropolitana de San Luis Potosí (ZMSLP) se localiza estratégicamente entre las ciudades más pobladas de México: la capital del país (México, D.F.), Guadalajara y Monterrey; además, es parte de la ruta de migración de personas provenientes de países centroamericanos hacia Estados Unidos. Esto ha ocasionado un vertiginoso crecimiento industrial y de servicios durante los últimos veinte años. Para conocer la susceptibilidad de urbanización, se desarrolló un modelo de regresión logística basado en variables espacialmente explícitas generadas a partir de imágenes de satélite Landsat (1990 y 2009) e información geoespacial. De las variables examinadas, la distancia a la zona industrial, la elevación y el costo del uso del suelo tuvieron un comportamiento estadísticamente significativo. Se presentan las zonas de susceptibilidad de crecimiento urbano como una herramienta de planeación territorial a nivel regional. El modelo estadístico fue validado satisfactoriamente mediante un algoritmo de validación cruzada con 10 iteraciones.

This paper presents an incremental k-most similar neighbor classifier, for mixed data and similarity functions that are not necessarily distances. The algorithm presented is suitable for processing large data sets, because it only stores in main memory the k most similar neighbors processed until step t, traversing only once the training data set. Several experiments with synthetic and real data are presented.