Search

En este Trabajo de Fin de Grado (TFG) se desarrolla una aplicación web de gestión de tareas potenciada por inteligencia artificial (IA), diseñada principalmente para equipos ágiles. La herramienta está construida para permitir a los usuarios y equipos ser más productivos con sus tareas, aprovechando el potencial de los LLM. En un mundo donde la inteligencia artificial está tomando un gran protagonismo, ignorar las posibilidades y ventajas que esta nos puede ofrecer en nuestras tareas del día a día, sería quedarse atrás ante un inminente cambio que inevitablemente, para bien o para mal, va a estar presente en muchos aspectos de nuestra vida. En el contexto de las metodologías ágiles, la gestión eficiente de tareas y la coordinación entre miembros del equipo son aspectos cruciales para el éxito de los proyectos. Este TFG presenta una solución innovadora que integra un asistente virtual basado en IA para ofrecer recomendaciones personalizadas y asistencia en tiempo real a los miembros del equipo. La principal motivación detrás de este proyecto es la necesidad de una herramienta que no solo facilite la gestión de tareas, sino que también potencie la productividad mediante el uso de inteligencia artificial. La falta de soluciones personalizadas y adaptativas en el mercado actual de aplicaciones de gestión de tareas para equipos ágiles impulsó el desarrollo de esta aplicación. Además, se busca explorar cómo los modelos de lenguaje grande pueden integrarse de manera efectiva en herramientas de software para ofrecer beneficios tangibles en el ámbito de la gestión de proyectos. Los objetivos principales de este TFG son: 1. Desarrollar una aplicación intuitiva y fácil de usar que permita a los usuarios gestionar sus tareas personales y de equipo de manera eficiente. 2. Integrar un modelo de IA dinámico que actúe como asistente personal, proporcionando recomendaciones y ayuda para mejorar la productividad basado en las tareas del usuario. 3. Evaluar el impacto de la integració

El Internet de las Cosas es un concepto basado en la conexión entre diferentes dispositivos muy extendido en la actualidad, que se encuentra presente en una gran diversidad de áreas como la industria, la salud o la domótica. Estos dispositivos son capaces de tomar datos del entorno y en caso de ser necesario, tomar una acción determinada, todo ello sin intervención de la persona. El presente trabajo de fin de grado se desarrolla bajo el contexto de los sistemas IoT en el ámbito de la salud. Este nace de la necesidad de monitorizar ciertos parámetros de pacientes de manera continuada. En este proyecto se pretende construir un sistema IoT capaz de tomar datos a través de una pulsera inteligente, la cual está conectada a una aplicación Android donde, además de poder consultar las mediciones, éstas serán preparadas para enviar a un servidor donde serán almacenadas, listas para ser empleados en cualquier proceso de análisis que sea requerido. A lo largo de este documento se describen tanto herramientas ampliamente utilizadas en el desarrollo de aplicaciones Android, plataformas empleadas en los sistemas IoT, o el estándar Fast Health Interoperability Resources, también conocido como FHIR y sus posibles implementaciones. ABSTRACT: The Internet of Things is a concept based on the connection between different devices that is currently widespread and is present in a wide range of areas such as industry, health or home automation. These devices are able to take data from the environment and if necessary, take a certain action, all without human intervention. This thesis is developed in the context of IoT systems in the field of health. It arises from the need to monitor certain patient parameters on a continuous basis. This project aims to build an IoT system capable of taking data through a smart bracelet, which is connected to an Android application where, in addition to being able to consult the measurements, these will be prepared to send to a server where they will be st

Este proyecto introduce una herramienta que facilita dos aspectos cruciales de la evaluación en la educación superior: la creación de plantillas de exámenes tipo test y la corrección automatizada de las mismas. Su primera función permite a los educadores elaborar plantillas adaptadas a las necesidades específicas de sus pruebas. La segunda y destacada función emplea tecnología de reconocimiento de imágenes para evaluar las respuestas de manera automática y precisa. Al integrar estas dos capacidades, la aplicación garantiza la asignación exacta de las respuestas a cada estudiante con identificación única, reduciendo los errores comunes en la calificación manual y agilizando el proceso evaluativo. La integración de esta aplicación en nuestro entorno educativo se ve facilitada por una interfaz de usuario intuitiva, lo que permite una rápida adaptación por parte de los docentes y el personal administrativo. Tras la corrección automatizada, la aplicación genera un archivo Excel que sirve como informe detallado de los resultados. Este archivo es una herramienta esencial tanto para los docentes como para la administración académica, ya que proporciona una visión clara y organizada del rendimiento de los estudiantes. Cada fila del archivo contiene el ID único del estudiante, seguido de sus respuestas a las preguntas del examen y la nota final obtenida. Esta estructura de datos no solo facilita una revisión rápida y eficiente de los resultados individuales, sino que también permite a los educadores realizar análisis adicionales, como tendencias de rendimiento o áreas que requieren atención adicional en la enseñanza. Abstract: This project introduces a tool that facilitates two crucial aspects of evaluation in higher education: the creation of test-type exam templates and their automated correction. Its first function allows educators to develop templates adapted to the specific needs of their tests. The second and prominent function employs image recognition technology to ev

Over the last few years, there have been several attempts to provide software tools for the development of federated learning (FL) models. However, both the complexity of the concept itself and the high entry barrier of these tools have meant that their adoption has been limited. Considering the related benefits, especially in terms of preserving data privacy, and the need for this type of solution in specific areas where data sharing is impossible, not only from a practical point of view but also from a legal and even ethical perspective, it is necessary to advance in solutions that allow its use to be democratised and its deployment to be extended. With this objective in mind, FLIP (Federated Learning Interactive Platform) has been developed as a comprehensive, easy-to-use fully functional web-based FL network management platform that eases and accelerates the usage of federated datasets by researchers in real scenarios. In this sense, FLIP has achieved a SUS score of 84.64, confirming a high level of perceived usability as expected. Taking this into account, FLIP can help increase the productivity and adoption of FL by a wider audience.

Federated learning initiatives in healthcare are being developed to collaboratively train predictive models without the need to centralize sensitive personal data. GenoMed4All is one such project, with the goal of connecting European clinical and –omics data repositories on rare diseases through a federated learning platform. Currently, the consortium faces the challenge of a lack of well-established international datasets and interoperability standards for federated learning applications on rare diseases. This paper presents our practical approach to select and implement a Common Data Model (CDM) suitable for the federated training of predictive models applied to the medical domain, during the initial design phase of our federated learning platform. We describe our selection process, composed of identifying the consortium’s needs, reviewing our functional and technical architecture specifications, and extracting a list of business requirements. We review the state of the art and evaluate three widely-used approaches (FHIR, OMOP and Phenopackets) based on a checklist of requirements and specifications. We discuss the pros and cons of each approach considering the use cases specific to our consortium as well as the generic issues of implementing a European federated learning healthcare platform. A list of lessons learned from the experience in our consortium is discussed, from the importance of establishing the proper communication channels for all stakeholders to technical aspects related to –omics data. For federated learning projects focused on secondary use of health data for predictive modeling, encompassing multiple data modalities, a phase of data model convergence is sorely needed to gather different data representations developed in the context of medical research, interoperability of clinical care software, imaging, and –omics analysis into a coherent, unified data model. Our work identifies this need and presents our experience and a list of actionable les

Precision cancer medicine suggests that better cancer treatments would be possible guiding therapies by tumor’s genomics alterations. This hypothesis boosted exome sequencing studies, collection of cancer variants databases and developing of statistical and Machine Learning-driven methods for alterations’ analysis. In order to extract relevant information from huge exome sequencing data, accurate methods to distinguish driver and neutral or passengers mutations are vital. Nevertheless, traditional variant classification methods have often low precision in favour of higher recall. Here, we propose several traditional Machine Learning and new Deep Learning techniques to finely classify driver somatic non-synonymous mutations based on a 70-features annotation, derived from medical and statistical tools. We collected and annotated a complete database containing driver and neutral alterations from various public data sources. Our framework, called Driver-Oriented Genomics Analysis (DrOGA), presents the best performances compared to individual and other ensemble methods on our data. Explainable Artificial Intelligence is used to provide visual and clinical explanation of the results, with a particular focus on the most relevant annotations. This analysis and the proposed tool, along with the collected database and the feature engineering pipeline suggested, can help the study of genomics alterations in human cancers allowing precision oncology targeted therapies based on personal data from next-generation sequencing.

Sentiment analysis is an active topic in Natural Language Processing (NLP). It has attracted a significant interest of research community due to the wide range of applications, including social-media, fake news spotting and interactive applications. In this paper, we present a novel approach for semi-automatic background creation and conspiracy classification. For this purpose, a complete framework including novel recurrent models is proposed. The BORJIS : Best algorithm foR Joint conspiracy and sarcasm detection has been tested on twitter-crawled data and It is composed by: (a) the crawler and labelling module, (b) the features vector extraction and (c) the conspiracy classifier. BORJIS was compared with up-to-date techniques and it showed a significant improvement (≥ 10% accuracy) when applied to diverse datasets.

Mobile apps analytics represent a core set in the mobile industry to extract relevant data with the aim of modeling user's behavior. Current solutions to detect in-app user's activity are usually based on a continuous app code modification schema, which implies high development efforts and a clear problem to implement changes without compromising the time to come back to the market or even with dependencies in the user's app updates. In this article, we analyze the suitability of aspect-oriented programming for providing a more efficient way to detect user's activity inside apps, which may lead to obtain user analytics. We propose an innovative approach that relies on an in-app solution based on the embedding of a specific library and a configuration file for setting up the events to be tracked in real time, without additional code changes in the app. Thus, this new schema will reduce the time and effort costs derived from the integration of third party trackers.

Analyzing a user's first impression of a website is essential for interface designers, since it is tightly related to their overall opinion of a site. In fact, this early evaluation affects user navigation behavior. Perceived usability and user interest (e.g. revisiting and recommending the site) are parameters influenced by first opinions. Thus, predicting the latter when creating a website is vital to ensure users’ acceptance. In this regard, web aesthetics is one of the most influential factors in this early perception. According to this, we propose the use of low-level image parameters for modeling web aesthetics in an objective manner, which is an innovative research field. Our model, obtained by applying a stepwise multiple regression algorithm, infers users’ first impression by analyzing three different visual characteristics of website screenshots –texture, luminance and color– which are directly derived from the MPEG-7 descriptors. The results obtained over three wide website datasets (composed by 415, 42 and 6 websites respectively) reveal a high correlation between low-level parameters and the users’ evaluation, thus allowing a more precise and objective prediction of users’ opinion than previous models which are based on other image characteristics with fewer predictors. Therefore, our model is meant to support a rapid assessment of websites in early stages of the design process, to maximize the likelihood of the users’ final approval.

Analyzing a user's first impression of a Web site is essential for interface designers, as it is tightly related to their overall opinion of a site. In fact, this early evaluation affects user navigation behavior. Perceived usability and user interest (e.g., revisiting and recommending the site) are parameters influenced by first opinions. Thus, predicting the latter when creating a Web site is vital to ensure users’ acceptance. In this regard, Web aesthetics is one of the most influential factors in this early perception. We propose the use of low-level image parameters for modeling Web aesthetics in an objective manner, which is an innovative research field. Our model, obtained by applying a stepwise multiple regression algorithm, infers a user's first impression by analyzing three different visual characteristics of Web site screenshots—texture, luminance, and color—which are directly derived from MPEG-7 descriptors. The results obtained over three wide Web site datasets (composed by 415, 42, and 6 Web sites, respectively) reveal a high correlation between low-level parameters and the users’ evaluation, thus allowing a more precise and objective prediction of users’ opinion than previous models that are based on other image characteristics with fewer predictors. Therefore, our model is meant to support a rapid assessment of Web sites in early stages of the design process to maximize the likelihood of the users’ final approval.

One of the advantages of social networks is the possibility to socialize and personalize the content created or shared by the users. In mobile social networks, where the devices have limited capabilities in terms of screen size and computing power, Multimedia Recommender Systems help to present the most relevant content to the users, depending on their tastes, relationships and profile. Previous recommender systems are not able to cope with the uncertainty of automated tagging and are knowledge domain dependant. In addition, the instantiation of a recommender in this domain should cope with problems arising from the collaborative filtering inherent nature (cold start, banana problem, large number of users to run, etc.). The solution presented in this paper addresses the abovementioned problems by proposing a hybrid image recommender system, which combines collaborative filtering (social techniques) with content-based techniques, leaving the user the liberty to give these processes a personal weight. It takes into account aesthetics and the formal characteristics of the images to overcome the problems of current techniques, improving the performance of existing systems to create a mobile social networks recommender with a high degree of adaptation to any kind of user.

Content personalisation is one of the main aims of the mobile media delivery business models, as a new way to improve the user’s experience. In broadcasting networks, the content is sent “one to many”, so a complete personalisation where the user may select any content is not possible. But using the mobile bidirectional return channel (e.g. UMTS connection) visual targeted advertising can be performed in a simple way: by off-line storing the advertisement for selectively replacing the normal broadcasted advertisement. In fact, these concepts provide powerful methods to increase the value of the service, mainly in mobile environments. In this article we present a novel intelligent content personalisation system for targeted advertising over mobile broadcasting networks and terminals, based on user profiling and clustering, as a new solution where the use of content personalisation represents the competitive advantage over traditional advertising.