Air quality forecasting is a critical component of environmental monitoring and urban decision-making, enabling public administrations to anticipate pollution episodes and implement timely mitigation strategies. Despite the advances in artificial intelligence (AI), selecting the most appropriate forecasting model remains challenging due to the variability across pollutants, prediction horizons, temporal resolutions, and modeling formulations. This paper presents an AI-driven decision support framework for urban air quality forecasting, evaluated through a real-world case study in Alicante (Spain). The proposed approach systematically analyzes multiple forecasting configurations, including daily and hourly resolutions, univariate and multivariate settings, and both recursive and direct strategies. The study compares classical machine learning models (K-Nearest Neighbors, Multi-Layer Perceptron, XGBoost, and Random Forest) with recurrent deep learning architectures (LSTM and xLSTM). A comprehensive evaluation is conducted across several atmospheric pollutants and forecasting horizons using complementary error metrics and non-parametric statistical tests. The results demonstrate that no single model consistently outperforms others across all scenarios. Instead, predictive performance is strongly dependent on the pollutant, forecasting horizon, and modeling formulation. While deep learning models show advantages in capturing complex temporal dependencies, classical machine learning techniques remain competitive, particularly in short-term forecasting tasks. These findings highlight the importance of configurable and evidence-based predictive modules within decision support systems. The paper advocates for a context-aware model selection strategy, where forecasting models are dynamically adapted to pollutant-specific and horizon-specific requirements rather than relying on a single global solution.

Concept lattices are central to Formal Concept Analysis (FCA), offering a structured and interpretable way to represent and analyze relationships between objects and attributes. However, producing clear and interactive lattice visualizations remains challenging, especially for larger or more complex datasets. Existing FCA tools often lack interactive features or rely on outdated technologies, limiting their accessibility and usability. In this paper, we introduce lattice.js, a lightweight JavaScript library designed for the interactive visualization of concept lattices. Built on top of the D3.js framework, lattice.js supports hierarchical layout computation using a heuristic adaptation of the Coffman–Graham algorithm combined with barycentric reordering to improve readability. The library also offers interactive exploration features such as zooming, filtering, node selection, and reduced labeling strategies. Additionally, it provides structural metric computation and supports exporting visualizations in multiple formats. We demonstrate the library’s functionality using a well-known FCA example, highlighting its accessibility and practical value for FCA applications.

Predictive Mutation Testing (PMT) has emerged as a promising technique for reducing the high computational cost of traditional mutation testing. This is achieved by using predictive models to estimate mutant behavior without actually requiring their execution. However, despite its growing interest, PMT is still a young and evolving field and, therefore, its experimental studies remain particularly vulnerable to various methodological threats that may compromise the validity, comparability, and reproducibility of results. This paper aims to identify and examine key methodological threats in prior PMT studies that not only affect their validity, but may also limit the technique’s ability to reach its full predictive potential. For each of the eight identified threats, spanning the main stages of the PMT workflow, we describe its nature, analyze how it has been addressed (or overlooked) in previous work, and offer a validation list of recommended practices. We further conduct an empirical validation to substantiate the impact of these threats. In addition, based on the gaps identified in this analysis, we outline several open challenges that remain unexplored in the field, such as the lack of standardization in dataset sharing, the prevalence of Java-centric studies with method-level operators, and the challenges posed by class imbalance and project heterogeneity. This work contributes to strengthening the methodological soundness of PMT research, promoting more meaningful cross-study comparisons, and encouraging the adoption of practices that foster reproducibility, scalability, and real-world applicability.

The rapid digital transformation and the ongoing paradigm shift towards Industry 4.0 have significantly augmented interconnectivity through the integration of sophisticated technologies within industrial networks, server infrastructure, and cloud environments. Consequently, these interconnected systems are increasingly susceptible to critical security vulnerabilities, including illicit access to confidential information and acts of sabotage. Recognizing the limitations of existing security systems, including intrusion detection systems and security information and event management systems, in addressing novel threats, we propose complex event processing (CEP) as an effective countermeasure. CEP enables the timely and real-time detection of cyberattacks within operational technology (OT) and the Internet of Things (IoT) environments, exhibiting adaptability to the resource constraints of these devices. This work presents a framework of a specific architecture tailored for OT and IoT, implementing CEP on Advanced RISC Machine (ARM) devices to improve the early identification of cyber intrusions, including its human-machine interface (HMI). The proposed framework operates by capturing network traffic and processing it through a CEP engine to automatically identify anomalous patterns. The framework’s efficacy is validated through its integration into Navantia’s integrated bridge system within a real-world deployment.

Modern cities increasingly rely on heterogeneous data sources and intelligent systems to improve urban management and quality of life. Among the many dimensions of urban intelligence, air quality monitoring and prediction stand out as a critical service given the well-documented impact of atmospheric pollutants on public health. This paper presents a framework that repurposes existing traffic camera infrastructures, combined with meteorological data, to forecast pollutant concentrations without requiring extensive historical air quality records. A comparative evaluation of six deep learning architectures, namely Multi-Layer Perceptron (MLP), Long Short-Term Memory (LSTM), Convolutional LSTM (CNNLSTM), LSTM with Attention, CNNLSTM with Attention, and Transformer, is conducted across two major Spanish cities, Madrid and Bilbao. Each architecture is evaluated over 10 independent runs with different random seeds, and statistical significance is assessed through Friedman tests followed by pairwise Wilcoxon signedrank tests with Holm-Bonferroni correction. The CNNLSTM and LSTM with Attention architectures are statistically the best-performing models overall, with no significant difference between them (mean RMSE of 5.754 and 5.672, respectively; p = 0.846). However, no single architecture dominates all scenarios: the Transformer achieves the best results for SO2, PM2.5, and PM10 in Bilbao, whereas the LSTM and LSTM with Attention form the statistically best group for O3. These findings highlight that pollutant-specific model selection, rather than a one-size-fits-all approach, is essential for operational short-term urban air quality forecasting.

Computational intelligence focuses on intelligent computer systems that mimic human nature and linguistic reasoning. Sentiment analysis is an area of considerable relevance within computational intelligence. Multimodal sentiment analysis is an extension of textual sentiment analysis, where the sentiments of people’s opinions are analysed by including multimedia content in addition to textual content. This mode of sentiment analysis faces multiple problems, as the sentiments of text and multimedia content may be contradictory. In addition, another added factor is the imbalance of the data that these problems suffer from in certain topics, which causes a problem when generating intelligent models. In this paper, we design a novel approach for multimodal sentiment analysis, proposing a new way of labelling tweets, not always prioritising polarized classes but using annotator confidence. Moreover, during this design, an information integration and fusion methodology is proposed for the construction of a metamodel that includes fuzzy logic to perform information weighting according to the confidence of the annotator. This proposal has been applied a public unbalanced dataset of tweets with text and images, with a large unbalance towards the negative class label. Applying the proposed fuzzy methodology, we reached a macro-F1 score of 0.493 for the negative class, 0.681 for the neutral class, and 0.832 for the positive class. The model obtains satisfactory performance since the individual image and text sentiment analysis results are worse, especially the negative class, which in initial image classification achieves an F1 score of 0.08.

As academic programmes in cybersecurity evolve to meet the demands of an increasingly complex and research-driven technological landscape, there is a growing need for pedagogical strategies that promote not only technical proficiency but also scholarly engagement. This study presents an instructional methodology that integrates research-based learning and peer assessment into a postgraduate cybersecurity course. Over two academic years, students engaged in the full research cycle by writing scientific short papers, conducting double-blind peer reviews, and participating in a simulated academic conference. The approach combined technical analysis of code vulnerabilities using both conventional SAST tools and generative AI systems, along with the development of academic skills such as scientific writing, critical evaluation, and oral communication. Survey results and evaluation scores from 35 students indicate high levels of perceived learning, engagement, and understanding of authentic academic practices. The successful submission of two student papers to an international doctoral consortium further highlights the academic relevance and impact of the initiative.

Predictive Mutation Testing (PMT) has been developed to alleviate the significant computational demands of traditional mutation testing by employing machine learning models to predict whether a test suite can detect artificially introduced code changes, known as mutants. Despite advances in predictive accuracy, the interpretability of these models has been limited, restricting insights into why specific predictions are made. This study investigates the application of explainability methods, specifically SHAP and LIME, to enhance the interpretability of PMT models. Using a comprehensive dataset of over 4 million mutants sourced from real-world Java projects, we analyze the contributions of various features to model predictions. Key findings indicate that features related to test execution, such as the number of times a mutation is executed or the number of tests from the test suite covering the mutation, play crucial roles in accurately detecting mutants. Additionally, code complexity correlates with higher mutant survival rates. The influence of the mutation operator highlights that different mutation types inherently vary in detection difficulty. This work not only demonstrates the effectiveness of SHAP and LIME in understanding model decisions but also provides actionable insights for improving test case design.

With the increasing adoption of AI in safety-critical applications within urban environments, the interpretability of these systems is paramount. This study explores the application of Explainable Artificial Intelligence (XAI) techniques to enhance transparency in audio-based detection of emergency vehicle sirens, a crucial component in urban sound management. Adopting methods such as SHAP (SHapley Additive exPlanations) values, Permutation Feature Importance, and model-specific feature scores, this research identifies key audio features, including mid-frequency spectral contrasts and targeted chroma components, which significantly help in distinguishing siren sounds among urban noise. The study examines various machine learning models, identifying K-Nearest Neighbors (KNN) and XGBoost as top performers; KNN excelled in class-specific precision, while XGBoost demonstrated strong cross-class discrimination. The findings highlight the potential of XAI in improving both accuracy and accountability for sound detection systems in safety-critical urban applications, advancing the deployment of transparent AI within smart city infrastructures.

The increasing adoption of conversational interfaces and semantic technologies in education is enabling new forms of scalable, personalised, and reflective assessment. In this context, a self-assessment platform has been designed and deployed within Telegram to promote formative learning through collaborative content creation and iterative practice. The system allows students to generate multiple-choice questions (MCQs), which are automatically classified into instructional topics using Natural Language Processing (NLP) techniques. These questions are reused to construct personalized quizzes, offering immediate feedback and supporting iterative improvement. During a semester-long deployment involving over 149 participants, students authored 519 MCQs and completed 1,709 quizzes. Results indicate high engagement, consistent score improvements across attempts, and effective reuse of learner-generated content, demonstrating the potential of NLP-assisted, mobile-friendly assessment to support self-regulated learning and performance growth.

Mutation testing evaluates test suite effectiveness by introducing small code modifications ("mutants") and checking if tests detect them. A major challenge is equivalent mutants, which are code changes that do not alter program behavior, making detection computationally expensive and undecidable. This review examines Machine Learning approaches, including Abstract Syntax Tree Neural Networks and Tree-Based Convolutional Neural Networks, to address this issue. Key challenges include limited datasets, inconsistent metrics, and reproducibility concerns. The study highlights the need for standardized frameworks and future research directions in mutation testing.

The ADEPT framework integrates Ambient Intelligence (AmI) technologies into Ambient Assisted Living (AAL) and Ubiquitous Computing to improve the quality of life for the elderly and those needing special care, particularly as populations in developed nations age. ADEPT addresses the complexities of data transmission and gathering within dynamic networks by utilizing edge computing for near-source data preprocessing, which enhances responsiveness and reduces network load. Its effectiveness is validated through simulations focusing on nursing home scenarios using the ns-3 network simulator and BonnMotion. The framework’s architecture facilitates efficient data handling by dynamically managing and prioritizing data flow through its network of nodes. Evaluations show that data prioritization significantly boosts data gathering success rates across different network setups, underscoring ADEPT’s potential to enhance data management in AmI applications and meet the changing needs of AAL environments.

This study evaluates the efficacy of Predictive Mutation Testing (PMT) models, focusing on the impact of preprocessing, class rebalancing, and model selection on predictive accuracy. By filtering unreached mutants, artificially inflated metrics caused by trivial correlations were minimized, enhancing the ability of the model to identify complex patterns. While ensemble models like CatBoost and XGBoost showed high effectiveness in detecting killed mutants, their performance in identifying survived mutants was limited, underscoring the need for refined strategies to address class imbalance. The study further identifies that previous PMT research often lacks class-specific metrics, such as the Matthews Correlation Coefficient, essential for assessing balanced performance. This oversight can yield misleading evaluations by favoring the majority class, which may falsely indicate high model reliability. Although SMOTE rebalancing improved recall for survived mutants, balancing sensitivity and precision remains challenging. The findings advocate for the development of more balanced datasets in future PMT research to enhance both accuracy and robustness in real-world testing scenarios.

Road safety is a critical global issue, with traffic accidents causing substantial human and economic losses annually. Advanced Driver Assistance Systems (ADAS) have emerged as an effective solution, leveraging technologies such as computer vision and artificial intelligence to enhance driving safety. However, the high costs and complexity of these systems limit their widespread adoption. This paper presents a cost-effective ADAS prototype utilizing vision-based and deep learning technologies to monitor and maintain safe following distances between vehicles. The performance of the system is evaluated using standard datasets and real-world scenarios, demonstrating adaptability across diverse driving conditions. Additionally, the feasibility of deploying the system on low-resource devices, such as Raspberry Pi and Radxa Zero, is analyzed, highlighting its potential for practical applications.

Pollution in urban areas is turning into a primary focus for local governments in developed nations around the globe. Lots of data are currently collected for this from smart developments provided with atmospheric and climatic sensors. A hot research line is now exploiting such data to extract patterns and predict pollution levels in such a way that countermeasures can be taken beforehand and exposure to harmful concentrations is avoided. However, a key issue is the lack of significant data, due to incomplete smart infrastructures or calibration problems in sensors. Dealing with this, in this paper we propose the exploitation of synthetic data generation to enhance pollution prediction based on limited data sources, concretely extending real measurements of two weeks to up to ten extra years. We present a data generation approach based on Generative Adversarial Networks (GANs), with a particular model focused on generating artificial pollution data, which is later exploited using different Machine Learning (ML) algorithms. Results indicate that the usage of synthetic data further improves prediction when used as the basis dataset to be later finetuned using real records. For 62% of pollutants this way to proceed in data mixing (among five different approaches) provides the best results in evaluations. Such effect is due to extra model robustness due to data regularization, and better generalization capabilities by avoiding sensor limitations in real deployments.

With increasing urbanization, efficient urban traffic management is a critical challenge that requires smarter and more adaptable systems. This paper introduces a self-learning algorithm designed to enhance the adaptability and effectiveness of vehicle detection models using urban camera infrastructures. By leveraging these ubiquitous devices, the study aims to capture and analyze real-time traffic data, a task traditionally limited by the need for extensive manual data labeling and the limitations of pre-trained models under varying urban conditions. Our self-learning algorithm addresses these challenges by reducing reliance on manual labeling and enabling continuous model adaptation to new conditions without direct human intervention. Implemented in the dynamic urban environment of the city of Madrid, Spain, this study evaluates the algorithm’s capacity to enhance vehicle detection, considering a diverse range of vehicle types. The core of the algorithm comprises an iterative self-training process that refines model performance using both labeled and unlabeled data, thus progressively enhancing detection accuracy. Our findings reveal significant improvements in the ability of the model to accurately identify and classify vehicles, highlighting the potential of self-learning algorithms in urban traffic management.

This paper presents a novel Multi-DL Fuzzy Approach aimed at performing image recognition in the development of a real-time traffic alert system, addressing the problem of traffic congestion and related incidents. Traditional monitoring by road operators predominantly relies on fixed location cameras, yielding limited and sometimes ambiguous information. This study proposes leveraging Twitter (now known as ‘X’) as a more comprehensive data source alongside employing fuzzy techniques with Deep Learning (DL) neural networks such as CNN, VGG16, and Xception to analyze and classify traffic images. The innovative integration of these technologies augments the precision in categorizing varying traffic conditions, namely fluid and dense traffic, accidents and fires. Thus, this proposal mitigates the ambiguities prevalent in traffic image interpretation, and reduces the dependency on static data sources. The proposed models showed improved results by combining information from the DL models, elevating accuracy from 84% in crisp classification to 90% utilizing fuzzy information.

Countries all around the world recognise the impact of air quality on public health, advocating for city centre decarbonisation and pollutant monitoring via Internet of Things technologies. Using data collected from these systems, it is possible to generate models that predict pollution based on regular patterns where mobility data can enhance the accuracy and robustness of these advanced machine learning models. This paper follows this approach, utilising vehicle traffic data from image recognition, on-site vehicle detectors, and synthetic data to maximise prediction accuracy in various urban environments. The results reveal that this proposal improves prediction for traffic-related pollutants, such as $SO_{2}$ and $PM_{2.5}$ , which are linked to severe respiratory diseases. These results also highlight the role of synthetic data in enhancing prediction performance under limited datasets.

This paper introduces the AL-MUSACTRA platform, an initiative designed to democratize access to cultural heritage through digital innovation, specifically catering to the diverse needs of individuals with disabilities. Grounded in the universal right to leisure and culture, the platform is a direct response to the mandates of international conventions and the Sustainable Development Goals, with a focus on inclusivity, safety, resilience, and sustainable access to cultural and natural heritage. Despite the progress in accessible audiovisual translation, the challenge of ensuring universal access to cultural heritage persists, particularly for those with visual, hearing, and cognitive disabilities. The AL-MUSACTRA platform leverages the Drupal Content Management System to overcome these barriers, offering features such as audio descriptions, sign language videos, and simplified texts. This paper details the platform's development, highlighting its emphasis on web accessibility standards and its potential to serve as a model for similar initiatives globally.