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

In today's world, the rapid progress of artificial intelligence (AI) and machine learning (ML) presents remarkable opportunities for developing innovative solutions to tackle various challenges within the healthcare sector. Deep learning (DL) has become a powerful tool in healthcare, transforming patient care and improving clinical support. It is increasingly utilized to identify critical features in medical images that go beyond what the human eye can naturally detect. Chest X-ray images are a widely used medical tool for detecting various health conditions. This covers pneumonia, lung cancer, and other issues such as tissue damage and bone fractures. Regardless of experience, for radiologists, accurately identifying diseases from X-ray images can be a strenuous task. Diagnosing pneumonia, a viral lung infection, is especially difficult because its symptoms closely resemble those of other pulmonary diseases. This similarity reduces the accuracy of current diagnostic methods. The vast amount of information contained in X-ray images has created an increasing demand for computerized support systems. This paper compares various computer-aided pneumonia identification methods, incorporating different deep learning approaches to streamline diagnosis using images of chest X-rays. In this study, seven types of deep convolutional neural networks have been applied to a dataset containing 5,856 Chest X-ray images of normal and pneumonia cases. It has been observed that VGG-16, VGG-19, and ResNet-50 effectively classify images of Chest X-ray into normal and pneumonia affected cases. Among these architectures, VGG-16 performs the best with an accuracy of 91%, followed by VGG-19 at 90.38% and ResNet-50 at 89.94%. The results surpass those of the advanced techniques mentioned in the literature.

Recommendation Systems (RS) are useful tools to help users find items of interest within an universe of options in this era of Big Data 3.0. Models such as Bias-SVD, SVD++, and their variants became classic and widely used in the heart of RS in e-commerce. However, these models do not consider popularity biases due to the Matthew effect present in the data structure, which leads to unfair recommendations. To address this problem, some researchers have proposed strategies that compensate for long-tail elements to try to increase the probability of their being recommended, other approaches use techniques such as the Zipf distribution to generate the predictions without prior knowledge of the data. However, these proposals have not been widely accepted because they do not consider user-item interactions in the training process. This paper presents a strategy that makes use of the Zipf distribution in the Bias-SVD model to incorporate popularity biases and improve the fairness of recommendations. Three variants of said model were implemented to show the validity of the strategy. The loss function uses Mean Square Error (MSE) and the error minimization is done using the ADAM algorithm. For the Experimental work, two MovieLens data sets with different distributions were used. The results show that it is possible to improve the fairness of recommendations by reducing the Matthew effect in the Bias-SVD model.

Technology has become a fundamental tool for addressing mental health issues such as anxiety and depression. This study examines the role of technological anthropomorphism in facilitating human-machine interaction through five key technologies: Tamagotchi, Real Chatbots (like ELIZA), Advanced Alexa, ChatGPT-4, and Holographic Interfaces. By evaluating these tools across factors such as emotional interactivity, personalization, and privacy, the results reveal that modern technologies like Advanced Alexa, ChatGPT-4, and holographic interfaces offer significantly enhanced capabilities, particularly in interpreting context and recognizing emotions. Holographic technology adds a unique dimension to this analysis by providing a three-dimensional, lifelike representation of the machine. This allows for more engaging and intimate interactions, which can be pivotal in creating more meaningful connections for individuals dealing with mental health challenges. The ability to see a hologram mimic human expressions and movements may foster a deeper emotional response, helping users feel understood and supported. This underscores the importance of ethical considerations in leveraging human-like interaction for mental health applications. A sample of 87 Generation Z participants (47 females and 40 males) from a Private University in Mexico was analyzed to investigate the interaction between these technologies and their potential in detecting anxiety and depression. This demographic provides unique insights into how digital natives perceive and interact with anthropomorphized technology.

This paper presents the development of an advanced clinical interface built on the LattePanda Sigma, an embedded device designed for edge computing. The interface integrates OpenAI language models and Whisper for automated speech-to-text transcription, together with accurate speaker diarisation in clinical settings using the pyannote/speaker-diarization-3.1 model. A dataset of ten doctor–patient conversations in Spanish—translated and re-recorded to suit the local context—was used to evaluate the models. Automatic transcriptions generated by the models were compared with the reference transcripts using the ROUGE metric. Average ROUGE scores of 0.9028 for the Small model and 0.9260 for the Medium model indicate high transcription accuracy. The reference transcripts were also used to assess the segments identified by the pyannote model. Finally, the paper analyses the system’s usefulness and effectiveness in improving Spanish-language clinical records.

This study presents a comprehensive dataset designed for the visual detection of crop diseases, comprising 43,267 images of 12 crop species across 15 disease classes. The dataset was developed over 14 months of dedicated human effort. To evaluate its effectiveness, several plant disease detection and classification algorithms were tested. The models generated by these algorithms were deployed on mobile devices and specialized hardware, enabling practical applications ranging from drones to Android smartphones, with on-device detection capabilities. The results highlight the performance of deep learning techniques, with the YOLOv4 algorithm achieving a mean average precision (mAP) of 71.04%, while the VGG model attained 92% precision and 90% accuracy. These findings demonstrate the potential of deep learning in enhancing crop monitoring, offering significant support for pest and disease control in vegetable crops. This work underscores the role of advanced technologies in promoting sustainable agricultural practices.

Medical simulators provide a safe environment for practising crucial procedures, particularly in virtual simulators where objective and quantitative data can be collected for developing machine learning algorithms for automatic expertise classification. This survey analyses 13 automatic evaluation systems used in medical simulators and identifies best practices for integrating ML algorithms. Among these systems, nine employed commercial simulators, particularly NeuroVR and the Da Vinci robotic systems, while four utilised custom simulators. The survey outlines the main steps in the integration of machine learning algorithms: data collection, metric generation and selection, training, and testing. Metric selection was identified as a crucial factor affecting both the accuracy of the algorithm and the comprehension of the evaluation. Typically, multiple machine learning algorithms were applied to the same dataset to compare results and identify the most effective model. Overall, this survey suggests that transparent algorithms are preferable, as they enhance physicians’ understanding.

In Mexico, the municipal marginalization index is used as a fundamental tool for planning and formulating public policies aimed at improving the socioeconomic conditions of the population. However, one of the limitations of this index is its exclusion of residential electricity consumption. In this regard, the present study explores the relationship between electricity consumption and the marginalization index. The research was conducted following the methodological framework of Data Science. As a result, geographic areas that cover municipalities with similar levels of residential electricity consumption were identified and characterized. It is noteworthy to mention that an inverse correlation was found between such consumption and marginalization index. Finally, these contributions could improve decision-making processes in public policies aimed at reducing economic and social inequalities, thus promoting more equitable development of the most lagging municipalities.

Novel feature selection methods are emerging to improve the accuracy of machine learning classifiers, including the method PowerSHAP (PS). This work analyzes the impact of PS to enhance the accuracy of Advanced Persistent Threat (APT) prediction in network traffic data. The dataset used in the experiments is DAPT2020, a labeled collection of network traffic data spanning Monday to Friday. The experimental data focuses on Wednesday’s traffic, which contains the majority of APT attack classes, such as Directory Bruteforce, Malware Download, Account Discovery, SQL Injection, CSRF, and the Normal class. Three experiments were conducted to assess the impact of feature selection with PowerSHAP in comparison to the standard data mining process.

Nonlinear systems have not been extensively utilized in engineering research. This is a consequence of the lack of effective analytical approaches analogous to those developed for linear systems, which are well understood and readily analyzable. This paper presents the application of two artificial neural networks (ANNs) for online identification of nonlinear systems as an alternative to linear systems theory. To illustrate this, the nonlinear model of a single inverted pendulum on a cart (SIPC) was identified from experimental data obtained from a laboratory prototype. The model structure employed in this study was based on the nonlinear autoregressive model with exogenous inputs (NARX) on Volterra polynomial basis function (VPBF) and Chebyshev polynomial basis function (CPBF) neural networks. The neural network structures were trained with experimental data from the SIPC prototype, which was recorded for a duration of 60 seconds. Subsequently, these models were validated using experimental data from a separate 15-second recording. The neural networks with Chebyshev polynomials demonstrated slightly superior performance.

This paper presents a systematic literature review of the application of retrieval-augmented generation (RAG) systems in educational settings, with a focus on teaching software engineering and related computing disciplines. Drawing on case studies, academic experiments, and surveys of teachers and students, it provides an overview of the current landscape, highlighting perceptions, reported effectiveness, and the technology’s impact in academia. Based on an analysis of 71 selected scientific papers, the review synthesises evidence on the extent to which RAG systems mitigate hallucinations and improve human–AI interaction. In addition, it suggests that many approaches discussed across studies could be strategically aligned with the integration of DevOps practices and RAG, enhancing their use through automation, continuous improvement, and the agile adoption of technologies within educational processes.