Sistema Regional de Información
en línea para Revistas Científicas de América Latina,
el Caribe, España y Portugal

 In this research an experimental software is developed for the classification, analysis and enrichment of indicators of compromise (IOCs), codenamed THREATMOSAIC. This software can import IOCs in bulk, classifying them according to whether they are IPv4, IPv6, URLs, MACs, e-mails, DNS domains or MD5, SHA1 and SHA256 hashes, sorting and sanitizing them in an effective and efficient way. All this is combined with the STIX2.1 standard, generating a directional graph enriched with information obtained from analysis through third-party REST APIs. Mainly information collected through services such as Virus Total, Abuse IPDB, IP Stack or Whois. Finally, the software allows sharing threat information in STIX2.1 format through the TAXII protocol via a server to which requests can be made from threat exchange platforms.

This study proposes a methodology for the classification of vocal pathologies by comparing voice signals with electroglottographic (EGG) signals. The segmentation of the voice signal into temporal components and its transformation into recurrence plots through intuitionistic fuzzy clustering provides input for a deep learning model to classify voices as healthy or pathological. The results obtained show that the Inception-v3 model, when using intuitionistic clustering, achieves superior accuracy — particularly with EGG signals — reaching a peak performance of 87.8%. Furthermore, the F1 score is 0.885 for EGG and 0.860 for speech, demonstrating better performance on EGG signals.

The University Course Timetabling Problem is regarded as one of the most significant administrative activities in academic institutions. It is framed as a combinatorial optimisation problem involving the scheduling of courses, students, lecturers, and classrooms, making it extremely complex. This work presents an analytical study of the Timetabling Problem, including its classifications, techniques, advantages, disadvantages, and a case study based on the model implemented at the Instituto Tecnológico de Ciudad Madero.

Social networks have become an important source of information in recent years, offering a platform for visualizing people's opinions on various industries and research topics. These opinions may be expressed in Spanish, providing fresh and updated insights that can be analyzed using computational approaches. This research aims to understand the sentiment conveyed in ideas or opinions within Spanish text. As a result, various methods are evaluated to identify the most effective approach. Computational techniques, especially Natural Language Processing (NLP), have become essential for automating this analysis. While traditional machine learning algorithms have been employed for sentiment analysis, the rise of large language models since 2017 has introduced significant challenges in assessing their impact on various NLP tasks. This paper presents a comparison between machine learning approaches and calibrated large language models for sentiment analysis in Spanish texts, measuring their performance. The calibration process consists of two stages: a coarse calibration using exploratory methods and a fine calibration that involves an algorithm for searching hyperparameter values. The evaluation process showed that the most effective machine learning approach combines unigrams and bigrams with the Bayes algorithm, along with exploratory parameter tuning and feature selection, achieving an accuracy of 72.72% and an F1 score of 72.81%. Furthermore, by applying LoRA, a technique that optimizes the fine-tuning of pre-trained models, it was found a best model that we call and store as twitter-xlm-roberta-base/SentUAM, which reached an accuracy of 71.99% and an F1 score of 71.78% in the sentiment analysis of Spanish texts.

The groupings of cardiac arrhythmias allow the identification of common patterns, distinctive characteristics, and similarities between different cases. In datasets where less common types of arrhythmias are identified, these grouping tools can better classify each subtype. This research was carried out on electrocardiogram records from a data set with more than 10,000 patients, previously labeled by cardiology specialists into 11 heart rhythms and grouped according to medical guidelines into four groups. A preliminary analysis of an ongoing project for detecting cardiac arrhythmias using unsupervised learning tools: clustering is presented. Feature selection was performed using filter tools, and the RR interval was extracted from the ECG records to be incorporated into the dataset under analysis as a new attribute. Internal validation metrics are used to check the quality of the selected clustering methods.

The growth of social networks as mass media has enabled faster and closer interaction between users, but it also presents challenges, such as the risk of spreading hate speech. Early detection of such harmful posts is critical. This article presents a methodology to create a unique corpus of Spanish-language comments collected from MisProfesores.com platform, covering all states in Mexico. This process resulted in a dataset of 18,000 unlabeled samples and 853 manually labeled samples. In addition to describing the corpus construction process, the results of the evaluation of different models trained with these data are presented, as well as their comparison with previous works for toxicity detection, highlighting the relevance of the Spanish corpus development for specific tasks. As a result, our Transformer-based model performed better than the state-of-the-art models in the binary toxicity classification, reaching a value of 0.9649 in accuracy and 0.9645 in F1 score.

This study examines communication challenges in distributed teams operating within the Scrum framework. To enhance communication effectiveness, Berlo’s model was adapted. An instrument was developed and validated by experts through surveys and interviews with professionals experienced in agile project management. The instrument was subsequently applied to 44 practitioners across seven countries, assessing its capacity to improve communication and optimise the flow of information between teams and stakeholders. The findings confirmed that the adaptation of Berlo’s model is effective in addressing communication challenges in distributed teams. In conclusion, the implementation of Berlo’s model in Scrum teams reinforces clarity and cohesion in communication, thereby contributing to the success of agile projects in global environments.

In this paper, we delve into the techniques used to extract information from medical texts written in Spanish. Our study focuses on fine-tuning Large Language Models by training them on Spanish medical texts and adjusting various hyperparameters. We also explore traditional machine learning algorithms like support vector machines, decision trees, and nearest neighbours. Our analysis aims to evaluate the tool's ability to identify entities and relationships between them. Our results show that the support vector machine algorithm outperformed Large Language Models in entity identification, achieving a 78.72% F1 score compared to 68.04%. However, Large Language Models demonstrated superior performance in relationship identification, achieving a 57.15% F1 score compared to 35.5% for machine learning algorithms.

One of the classic combinatorial optimization problems belonging to the NP-hard class is the quadratic assignment problem. The interest in solving the problem lies in its high computational complexity, as well as its applications in: logistics, electronic circuits, gate assignment in airports, among others. In this work, the Greedy Random Adaptive Search Procedure metaheuristic was implemented to find its solutions. The main contribution of this work is the adaptation of a neighborhood structure contained in k-exchanges in the post-processing phase. The tests were performed for 29 large-scale instances whose dimensions range from 64 to 254 taken from the QAPLIB. The approximate solutions were found through a metaheuristic that bases its search on neighborhoods and local search algorithms. Java was the programming language used for the implementation of metaheuristics; its execution allowed balancing the parameters to obtain competitive results with respect to the values ​​known in the literature. The results reported and presented in this work achieved the proposed objectives.

This work presents an autonomous robotic platform aimed at improving indoor air quality and ensuring effective disinfection. The system leverages UVC germicidal radiation and a nebulization module to eliminate pollutants, including ozone and microbial pathogens, aligning with Sustainable Development Goals (SDGs) 3 and 11. The robotic platform incorporates six 253.7 nm UVC lamps arranged to maximize disinfection efficiency, achieving 95% microbial reduction within 14.76 seconds for irradiated areas of 0.98–1.04 m². An integrated olfactory system, using MQ-131 and MQ-135 sensors, monitors ozone and CO2 concentrations, ensuring environmental safety by automatically shutting down disinfection processes when thresholds are exceeded. The nebulization system disperses a hydrogen peroxide (H2O2) solution, reducing ozone levels from 29.22 ppm to 0.05 ppm in enclosed spaces within 3 seconds. The experimental results demonstrate the platform's capability to sanitize surfaces, walls, floors, and ceilings in diverse indoor environments, such as hospitals and classrooms, with minimal human intervention. By combining tele controlled and autonomous functionalities, the system offers a scalable solution for sustainable and inclusive urban environments, addressing key challenges in air quality and health resilience post-COVID-19. This innovation contributes significantly to the goals of creating healthier living spaces and advancing the development of smart, sustainable cities.