Amidst growing environmental concerns and the push for sustainability, the intricacies of energy management in diverse building settings demand innovative solutions. Traditional strategies fail to address the unique characteristics and needs of different building types and users leading to inefficiencies and a lack of optimization in energy use. This study presents a sophisticated building energy management system, assessed via pilot implementations, which utilizes a platform based on adaptive and intelligent edge computing. This comprehensive system integrates various elements such as smart meters, HVAC controls, electric vehicle charging stations, and smart plugs. Utilizing the ICT PSP framework, the methodology encompasses rigorous data collection and analysis, emphasizing usability, error rectification, and the detailed monitoring of impacts. Key findings reveal that tailored energy management solutions, augmented by real-time data and user-centric interfaces, significantly improve energy efficiency. The study also highlighted the impact of socio-economic factors and weather conditions on energy consumption, underscoring the necessity of incorporating these variables into energy management strategies. The integration of demand-side management (DSM) through energy retailers, distribution system operators (DSO), or other power grid stakeholders has yielded valuable insights into external factors affecting energy consumption patterns. Specifically, these insights pertain to electricity tariffs and consumer behavior.
In the intricate process of shipbuilding, digitalization plays a significant role in enhancing competitiveness and sustainability. This multifaceted process involves numerous minor pre-assemblies, where automation, specifically through robotic welding, proves instrumental in enhancing overall efficiency. This paper utilizes real data from a shipyard to implement an artificial vision system designed to identify the elements required for assembling a minor pre-assembly. Surface-based matching and 3D edge matching techniques were applied to a diverse dataset of point cloud images. By employing multiple computer-aided design models to generate corresponding surface models, we successfully identified all the pieces present on a tray. This identification process is particularly complex due to the multitude of minor pre-assembly types, the challenge of distinguishing between similar components, and the stacking of these parts. The distinction of these elements significantly contributes to ensuring traceability throughout the hole shipbuilding process.
Measurement in quantum theory is a key scientific case where one of many possible outcomes comes to pass, the other possible outcomes disappear, and the world continues from a newly-fixed state. In quantum theory, Lüder’s projection postulate constraints the way the state of a system has to change following measurement, an idea which has been multiply employed in applications of quantum theory outside physics, notably in cognition. Recently, Orrell presented an economic model based on quantum theory. We summarise this view, including evidence for its plausibility. Following from Orrell’s ideas, we consider how the notion of a system, measurement, and Lüder’s postulate might translate in a financial context. Specifically, quantum theory predicts a so-called quantum Zeno effect, that is, a reduction that some initial state changes, after receiving various pieces of information all pushing towards change, as the density of intermediate measurements increases. We aim to translate this idea to the financial world and we believe there is an application in the real estate market: Due to the individual nature of houses and the corresponding scarcity of comparable properties, we observe especially pronounced relationships between low liquidity and price uncertainty in the property market. We propose that this way of thinking essentially links low liquidity with volatility and we offer some preliminary analyses substantiating our point.
In a constantly expanding global market, where worldwide container traffic surged nearly fourfold between 2000 and 2019, predicting demand becomes crucial for a company’s optimal growth. Particularly, the fashion sector is highly unpredictable, characterized by: seasonality, cultural influences, and fashion trends. These factors make it challenging to forecast footwear demand between seasons. In recent years, various algorithms for demand prediction have been explored, generally classified into three main categories: statistical, artificial intelligence and hybrid algorithms, each with its unique characteristics. The goal of this work is to predict the sales of a specific shoe model for a company. To achieve this, some Key Performance Indicators (KPIs) have been established that provide a comprehensive description of the prediction made. Holt-Winters has been chosen as the prediction algorithm, which achieves an accuracy rate of 91.6
The emergence of Industry 4.0 as a de facto paradigm for the industries of the future still presents some needed innovation and problem solving. In a increasing interconnected industrial setup where all nodes in a industrial process are increasingly interconnected, new challenges arise from the need to gather and explain machine learning models obtained from data streams present in multiple locations. These problems may be tackled with the fields of explainable artificial intelligence (xAI) methods and federated learning. The data constraints in temporal dependencies as is the case with most industrial process, may also make these problems fall under the timeseries category. This article presents an initial approach to deal with these problems and provides a theoretical overview of a federated xAI system for timeseries applications in Industry 4.0. The main objective is to present and discuss the integration of previous efforts into a large scale application for machine learning algorithm within I4.0 that offer human understandable decisions for I4.0 decision makers.
The Zigbee protocol, designed for low-power personal area wireless networks, is a technology widely used on the Internet of Things. This paper presents a study on the detection of denial-of-service attacks in Zigbee networks using supervised classification algorithms. Three techniques are evaluated: Logistic Regression, K-Nearest Neighbors and Support Vector Machines. A generated dataset is used for the analysis, and the results show that the K-Nearest Neighbors and Support Vector Machines approach achieves high performance and low computational demand. This methodology offers a promising strategy for security in Zigbee networks.
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.
Parkinson’s Disease ranks as the second most prevalent neurodegenerative disorder globally, second only to Alzheimer’s Disease. Its impact is significant, affecting an estimated 7 to 10 million individuals worldwide. Notably, its incidence rises with age, typically manifesting after 50 years old. As the global population ages, the prevalence of Parkinson’s Disease is projected to escalate proportionately, presenting a considerable public health challenge. Presently, diagnosing Parkinson’s Disease remains challenging, lacking a definitive and universally applicable method. Therefore, there’s a growing interest in exploring alternative approaches, such as leveraging Machine Learning (ML) algorithms, particularly those trained on voice datasets, to enable early detection. Initial examination of available datasets reveals inherent imbalances, underscoring the need for meticulous preprocessing steps to ensure accurate and reliable analysis. In this study, a comprehensive investigation into various ML algorithms is proposed, incorporating a range of preprocessing techniques tailored to address dataset complexities. Notably, a Hybrid Classification System is suggested, integrating SMOTE to mitigate imbalance, feature selection algorithms to enhance predictive accuracy, and Ensemble methods to amalgamate diverse classifiers’ outputs, thereby optimizing diagnostic performance, achieving 98.3
Multiple open data portals offer data that may not appear to violate data privacy or confidentiality laws at first glance. However, a thorough study of these datasets and their relationships with others reveals that confidential or private information may be obtained in certain cases. To address these issues, this article proposes a solution that involves implementing a series of AI-powered modules. The goal of these modules is to analyze the quality of the data and its potential combinations with linked data that could lead to legal non-compliance or data quality issues. Due to the lack of standardization across different open data portals, this model facilitates the improvement of these portals for information extraction and decision-making purposes while ensuring compliance with data privacy and confidentiality laws.
At the end of 2022, the Council of the EU approved the Sustainability Reporting Directive (CSRD), requiring companies to provide concrete detailed information, which strengthens company accountability, avoiding divergences between the different existing reporting frameworks and facilitating a transition to a sustainable economy, aligned with the Sustainable Development Goals of the 2030 Agenda within the framework of the European Green Deal and the Sustainable Finance Program. The most relevant aspect derived from the Directive is the application of the concept of “double materiality”, which focuses on the main impacts (triple impact) of companies in terms of sustainability (economic, social and environmental), constituting the strategic pillars of the ESG (environmental, social and corporate governance) strategic plans of organizations. Although the Directives will come into force from 2025, large European companies have begun to prepare sustainability reports focused on the definition of double materiality and its triple impact. This new paradigm will turn organizations into hubs in charge of regenerating social welfare, identifying financial opportunities with sustainability. In this paper, we undertake an analysis of the materiality matrices prepared by the large Spanish companies of the General Index of the Madrid Stock Exchange (IGBM), which stand out for the transparency and relevance of their information based on the Reporta 2024 Report in order to identify sustainable aspects in today’s society and economy.
The advancement of deep learning techniques has significantly improved diseases detection on plants remote sensing images. However, this imaging process comes with challenges due to the complex nature of plants remote sensing images abnormalities. The vast range of anomalies in terms of kind, shape, and magnitude of lesions makes it challenging to effectively detect them in different situations. However, differentiating an early-stage disease from a healthy plant on images is a major challenge, even for experienced professionals. Simultaneous identification of plant diseases in the same region re-mains a challenge for remote sensing imagery. In this paper, we build on the fact that several disease types exhibit a scale-sensitive feature that can be exploited by deep learning models based on multi-level features. We therefore propose to use the RT-DETR backbone in the YOLOv8 network for cotton disease detection. The integration of the RT-DETR backbone within the deep learning framework is to improve the detection and differentiation of cotton diseases and thus increase the capabilities of aerial imagery. This study analyzes five disease types: fungal leaf, asymptomatic leaf, viral leaf, co-infection, and boll rot. Experimentation with the model showed favorable results for detecting cotton diseases on UAV image by Deep Learning. The fungal leaf and boll rot classes showed the greatest results, with an intersection over union (IoU) of 0.7 and average accuracy of 82.7
Smart homes for ambient assisted living (SHAAL) can provide cognitive assistance and telemonitoring to foster independent living at home. However, building a personalized SHAAL is a complex and time-consuming co-construction process involving IT specialists, healthcare specialists, caregivers, and the elderly person herself. Not surprisingly, it is very difficult to access and integrate all these expertise at the same time in the same place. That is why we are developing a do-it-yourself (DIY) approach aiming at enabling a non-expert to build her own SHAAL. To get there, four phases need to be supported: design, installation, tests, operation, and maintenance. This paper focusses on the design phase. It presents an innovative prototype platform that leverages Augmented Reality (AR) and Artificial Intelligence (AI) to guide a non-expert through a user-friendly “do-it-yourself” (DIY) process to design in situ a SHAAL. This platform captures each category of expertise using templates and models. Thanks to ontologies and case-based reasoning, AR makes available this expertise and assists a user while she is designing in-situ her SHAAL. Two preliminary experiments involving expert and non-expert users showed promising results.
Many of today’s domains of application of Machine Learning (ML) are dynamic in the sense that data and their patterns change over time. This has a significant impact in the ML lifecycle and operations, requiring frequent model (re-)training, or other strategies to deal with outdated models and data. This need for dynamic and responsive solutions also has an impact on the use of computational resources and, consequently, on sustainability indicators. This paper proposes an approach in line with the concept of Frugal AI, whose main aim is to minimize the resources and time spent on training models by re-using models from a pool of past models, when appropriate. Specifically, we present and validate a methodology for similarity-based model selection in data streaming environments with concept drift. Rather than training a new model for each new block of data, this methodology considers a pool with only a subset of the models and, for each new block of data, will select the best model from the pool. The best model is determined based on the distance between its training data and the current block of data. Distance is calculated based on a set of meta-features that characterizes the data, and on the Bray-Curtis distance. We show that it is possible to reuse previous models using this methodology, leading to potentially significant saving of resources and time, while maintaining predictive quality.
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 systematic review examines the use of heat maps for weapon detection, an important aspect given the growing need for security. A comprehensive search was conducted in academic databases, resulting in the retrieval of 35 relevant articles published between 2017 and 2023. The findings indicate increasing interest in the field, especially in 2023 as new research continues to emerge. Key technologies utilized include deep learning, thermal image processing, and object detection. These works provide valuable tools associated with their real-world implementation. The recommendations for future work include improving evaluation methods, incorporating more scenarios and diverse users for testing, and optimizing algorithms. This review highlights the potential importance of heat maps in enhancing weapon detection systems and security. This emerging field, which combines computer vision, deep learning, and thermal imaging, identifies opportunities for future research.
With the development of communications technologies, the remarkable progress of IoT and the growth of Industry 4.0 in recent years, research in cybersecurity is of increasing interest in industry and economic field. Moreover, quantum computing threatens to break traditional cryptographic systems and traditional cybersecurity schemes. Therefore, it is essential to deploy new solutions in order to overcome this challenge. In this paper we address some of the most relevant aspects of the field of cybersecurity in the context of the development of quantum technologies, both in terms of challenges to be overcome and new opportunities. In this sense, the article is divided into several sections that address the different aspects. Each of the sections consists of two parts. The first part consists of a brief bibliographical review. The second part introduces a framework and a roadmap to advance in the investigation of each of them in the future.
Nowadays, smartphones, tablets and car user interfaces are full of various applications that enable us to perform a wide variety of tasks. To increase comfort but also safety in some specific contexts such as driving, it is relevant to determine at any instant which is the next application that the user will use to facilitate the access to the predicted application. But if a user started very recently to use the device, we have very little data about him or her. Therefore, the Machine Learning (ML) model used to suggest the next application could quickly overfit and its suggestions would be inaccurate. To mitigate this problem, we propose an approach based on Data Augmentation (DA) using a variant of Generative Adversarial Networks (GAN) called DoppelGANger. Our ML model used to predict the next application is a Deep Neural Network (DNN) based on Long-Short term memory (LSTM) units. By adding the synthetic data generated by the GAN to the original training data, we obtain a mean improvement of 4.7
This article details the development and deployment of an Internet of Things (IoT) platform aimed at optimizing water usage in community kitchens in Cali, Colombia. Conducted at the Fundación Aprender, Crear y Crecer, which serves as both a dining hall for 120 people and a residence for the project leader’s family, this initiative employs an IoT architecture enhanced by Edge Computing technologies. This system integrates continuous data collection through IoT sensors, edge processing for predictive analysis, and a cloud platform for advanced data analytics and visualization, optimizing decision-making in real-time. The research findings indicate a strong positive correlation (Pearson coefficient approx. 0.926) between rising average temperatures and increased water consumption, especially during summer months. Data gathered over two years enabled the identification of consumption patterns and hypothesis formulation. Based on these insights, a scalable platform is proposed for implementation across 762 other community dining facilities in the city, offering a robust solution for large-scale water management. This strategy not only aims to enhance environmental and economic sustainability but also seeks to improve the quality of life for vulnerable communities through efficient water resource management.
Ground-based telescopes face a significant challenge posed by atmospheric turbulence, resulting in acquired images appearing distorted and lacking sharpness. Adaptive optics technology is employed to mitigate this issue by effectively correcting wavefront aberrations through adjustments to the surface of a deformable mirror. Furthermore, the integration of neural networks into the control system has demonstrated notable enhancements in both atmospheric correction and turbulence prediction. Specifically, in this work, a 2D-LSTM network structure is utilized, which has shown good efficiency in slope prediction over a time sequence. The objective of this study is to address the prediction of turbulence data without the noise introduced by reading instruments. Through the experiments conducted in this research, it is demonstrated that such neural models are capable of learning to a certain extent the noise patterns of the system. Thus, the obtained data closely resemble real-world turbulence conditions.
In today’s digital era, the proliferation of counterfeit websites poses a significant concern, as they aim to trick users into divulging personal and financial details. This study delves into the efficacy of Machine Learning methodologies, such as Extra-Trees Classifier, Extreme Gradient Boosting (XGBoost), and Decision Tree, in identifying fraudulent websites. We trained and assessed the algorithms by extracting various aspects from the content and metadata of websites that included malware, phishing, defacement, and benign websites using a categorical dataset. The findings reveal that the Extra-Trees Classifier yielded the highest accuracy rate (97
