Search

La seqüenciació d’RNA de cèl·lules aïllades (scRNA-seq) és un àmbit de recerca crucial per a la comprensió de la diversitat cel·lular, amb grans implicacions per a la recerca en biomedicina. Entendre la informació dispersa, sorollosa i complexa que contenen les dades del scRNA-seq és un problema fonamental de la biologia computacional. Aquesta tesi pren com a punt de partida un model probabilístic desenvolupat per Shiragur i Uhler que formalitza la generació de dades d’scRNA-seq. Basant-nos en aquest model, presented tres nous algoritmes d'identificació de tipus cel·lular, un dels quals implementa un enfocament novedós tot fent servir tècniques de descomposició tensorial per estimar els paràmetres del nostre model probabilístic i, seguidament, utilitzent aquests paràmetres per identificar amb precisió el tipus cel·lular. Els algoritmes que hem desenvolupat han estat validat mitjançant nombroses simulacions i anàlisis amb dades reals i mostren un rendiment comparable als mètodes més avançats. Per tant, ofereixen una sòlida base teòrica per a futurs desenvolupaments en l’anàlisi de l’scRNA-seq. Les simulacions demostren que aquests algoritmes són capaços d’estimar amb fiabilitat els paràmetres del model, fins i tot en el règim de baixa profunditat de seqüenciació, en què els models anteriors basats en clusterització no ofereixen bons resultats., La secuenciación de ARN de células individuales (scRNA-seq) es un campo de investigación crucial para comprender la diversidad celular y tiene importantes implicaciones para la investigación biomédica. Entender la información dispersa, ruidosa y compleja en los datos de scRNA-seq es un problema fundamental en la biología computacional. Este trabajo se basa en un modelo probabilístico desarrollado por Shiragur y Uhler, que formaliza el proceso de generación de datos de scRNA-seq. Aprovechando este modelo, presentamos tres algoritmos de identificación de tipos celulares, uno de los cuales implementa un enfoque novedoso utilizando técnicas de descomposición tensorial para estimar los parámetros de nuestro modelo probabilístico y luego usar los parámetros estimados para identificar con precisión el tipo de cada célula. Nuestros algoritmos, validados a través de extensas simulaciones y análisis de datos reales, demuestran un rendimiento comparable a los métodos más avanzados, proporcionando una sólida base teórica para futuros avances en el análisis de scRNA-seq. Nuestras simulaciones demuestran que nuestros métodos estiman de manera confiable los parámetros del modelo incluso en el régimen de baja profundidad de secuenciación, donde los enfoques previos basados en la agrupación no son suficientes., The rapidly growing field of single-cell RNA sequencing (scRNA-seq) is crucial for understanding cellular diversity and has significant implications for biomedical research. Understanding the sparse, noisy, and complex information in scRNA-seq data is a fundamental problem in computational biology. This work builds on a probabilistic model developed by Shiragur and Uhler, which formalizes the scRNA-seq data-generating process. Leveraging this model, we present three cell-type identification algorithms, one of which implements a novel approach by employing tensor decomposition techniques to estimate the parameters of our probabilistic model and then using the estimated parameters to accurately identify each cell's type. Our algorithms, validated through extensive simulations and real data analyses, demonstrate performance comparable to state-of-the-art methods, providing a solid theoretical foundation for future advancements in single-cell RNA-seq analysis. Our simulations demonstrate that our methods reliably estimate the model's parameters even in the low sequencing depth regime, where the prior clustering-based approaches fall short., Outgoing

The main goal of this thesis is to apply advanced intelligent segmentation techniques into a real case study from the cultural industries in Catalonia to successfully differentiate between types of customers by applying advanced clustering techniques, based on their characteristics, consuming behaviours along with others. The study investigates the integration of clustering methodologies to optimize event management strategies. This was materialized by applying a hierarchical clustering process like Clustering based on ontologies. Looking at patterns, trends and what people like in these events, combined with characteristics like types of events, who is attending, how engaged they are and more, will be a key aspect to figure out how to group similar customers together, understanding how people behave, and making better decisions when planning and running them. The introduction of ontologies as a semantic bias of the hierarchical clustering processes allows to guarantee a higher explainability capacity of the results of the clustering, so that the resulting clusters can be easily interpreted and understood by the stakeholders. The real application supported in this thesis seeks the clusters based in data about product consumption of the customers and the set of products available, formalized through a specific-domain ontology of products that was built as one of the outcomes of this thesis. At a methodological point of view, the study addresses challenges inherent in clustering analyses, such as deep preprocessing of input data, determining optimal cluster numbers and mitigating the impact of outliers and noise in event-related datasets. Advanced evaluation metrics and visualization techniques are employed to validate the robustness and reliability of clustering results. The application of explainable AI techniques help to automatically interpret the meaning of clusters to enhance the decision support potentiality of the results. In this case, automatic interpretatio

This is the peer reviewed version of the following article: Fonseca, P. & Ruiz, C., 2024. Building models in pairs for cross-verification using SDL and DEVS. Advanced theory and simulations. Available at: http://hdl.handle.net/2117/406231, which has been published in final form at https://onlinelibrary.wiley.com/doi/10.1002/adts.20230083. This article may be used for non-commercial purposes in accordance with Wiley Terms and Conditions for Use of Self-Archived Versions. This article may not be enhanced, enriched or otherwise transformed into a derivative work, without express permission from Wiley or by statutory rights under applicable legislation. Copyright notices must not be removed, obscured or modified. The article must be linked to Wiley’s version of record on Wiley Online Library and any embedding, framing or otherwise making available the article or pages thereof by third parties from platforms, services and websites other than Wiley Online Library must be prohibited., The objective of the paper is to present a methodology that can be used to translate a model from one formalism to another allowing model reuse and cross-verification. With the use of formal languages, the model specifications can be expressed in a rigorous and univocal way, and the model can be validated against the specifications or conceptual model. Expressing the same model in different formal languages opens the conceptual model validation to a varied number of specialists. Also, in the context of modeling a new system, where there is no data to perform validation against a real system, having pairs of models can be used to perform cross-model verification to ensure that the model specifications or conceptual models are correctly implemented by comparing the results produced by both models. Specifically, a method is presented to translate a model conceptualized on specification and description language (SDL) to discrete event system specification (DEVS). The transformation mechanism between SDL and DEVS formalisms is described. The methodology is exemplified with a disease spread model for COVID-19, and it is shown how the results obtained by the two models can be used for cross-verification., Peer Reviewed, Objectius de Desenvolupament Sostenible::3 - Salut i Benestar, Objectius de Desenvolupament Sostenible::3 - Salut i Benestar::3.d - Reforçar la capacitat de tots els països, en particular els països en desenvolupament, en matèria d’alerta primerenca, reducció de riscos i gestió dels riscos per a la salut nacional i mundial, Postprint (published version)

Els models climàtics són normalment descrits per una ecuació diferencial la solució de la qual és l'evolució temporal d'alguna magnitud que es vol estudiar. Com els mètodes clàssic poden ser cars d'implementar, introduim noves idees sobre com aproximar aquestes solucions amb l'ajuda de models d'aprenentatge profund. La idea principal és fer predicció sobre objectes amb poca resolució, deixant que diverses arquitectures d'aprenentatge profund aprenguin aspectes clau sobre la evolució i la distribució d'aquestes magnituds., Los modelos climáticos son usualmente descritos por una ecuación diferencial cuya solución es la evolución temporal de alguna magnitud que se quiere estudiar. Puesto que los métodos clásicos pueden ser caros de calcular, introducimos nuevas ideas de cómo aproximar estas soluciones con la ayuda de modelos de aprendizaje profundo. La idea principal es de hacer predicción sobre objetos con baja resolución, dejando que diferentes arquitecturas de aprendizaje profundo aprendan aspectos clave sobre la evolución y la distribución de estas magnitudes., Climate models are often described by a differential equation whose solution is the temporal evolution of a given magnitude we want to study. As classical methods can be quite expensive to compute we introduce new ideas on how to approximate these solutions with the help of deep learning models. The core idea is to perform forecasting on samples with low resolution by letting different architectures learn key features about the evolution and distribution of these samples, Outgoing

Els catxalots tenen un llenguatge divers i complex que actualment és investigat per grans experts amb l'objectiu de desxifrar el significat de les seves comunicacions. No obstant això, fins ara la investigació s'ha vist limitada pel fet que les dades disponibles són limitades. Fins i tot amb accés a la capacitat de col·locar micròfons a les balenes i gravar llargs arxius d'àudio que inclouen converses, l'adquisició de dades continua sent un coll d'ampolla en el procés per la necessitat de transcriure les comunicacions dins de làudio, que és un procés llarg i lent fins i tot per als biòlegs marins experts. En aquest treball presentem un mètode per fer aquesta transcripció de forma automàtica. Mitjançant la creació d'un procés fiable i ràpid per detectar les comunicacions de les balenes, esperem trencar el coll d'ampolla en l'estudi del llenguatge dels catxalots, facilitant el desenvolupament de la nostra comprensió. Les nostres contribucions són les següents: En primer lloc, presentem un mètode de detecció automàtica de clics de comunicació de catxalots dins d'àudio. Segon, mostrem com aquest mètode pot ser adaptat i continuat per agrupar aquests clics a codes. Finalment, mostrem com aquest mètode pot ser adaptat i continuat per identificar el parlant de les codes (codes son sequencies de clics dit per catxalots per comunicar)., Los cachalotes tienen un lenguaje diverso y complejo que actualmente está siendo investigado por grandes expertos con el objetivo de descifrar el significado de sus comunicaciones. Sin embargo, hasta ahora la investigación se ha visto limitada por el hecho de que los datos disponibles son limitados. Incluso con la posibilidad de colocar micrófonos en las ballenas y grabar largos archivos de audio que incluyen conversaciones, la adquisición de datos sigue siendo un cuello de botella en el proceso debido a la necesidad de transcribir las comunicaciones dentro del audio, que es un proceso largo y lento incluso para los biólogos marinos expertos. En este trabajo presentamos un método para realizar esta transcripción de forma automática. Mediante la creación de un proceso fiable y rápido para detectar las comunicaciones de las ballenas, esperamos romper el cuello de botella en el estudio del lenguaje de los cachalotes, facilitando el desarrollo de nuestra comprensión. Nuestras contribuciones son las siguientes: En primer lugar, presentamos un método para detección automática de clics de comunicación de cachalotes dentro de audio. Segundo, mostramos cómo este método puede ser adaptado y continuado para agrupar esos clics en codas (codas son secuencias de clics que las ballenas usan para comunicarse). Finalmente, mostramos cómo este método puede ser adaptado y continuado para identificar el hablante de las codas detectadas., Sperm whales have a rich and complex language that is currently being researched by top experts trying to decipher the meaning of their communications. However, research so far has been limited by the fact that the available data is limited. Even with the ability to place microphones on whales and record long audio files that include conversations, data acquisition is still bottlenecked by the need to transcribe the communications within the audio, which is a long and time-consuming process even for expert marine biologists. In this work, we present a pipeline for automatically doing this transcription. By creating an accurate and fast pipeline for detecting whale communications, we hope to break the bottleneck in studying sperm whale language, making it much easier for our understanding to develop. Our contributions are the following: First, we present a method for automatic detection of sperm whale communication clicks within audio. Second, we show how that method can be adapted and continued to group these communication clicks into codas (codas are sequences of clicks that whales use to communicate). Finally, we show how that method can be adapted and continued further to identify the speaker of the detected codas., Outgoing

Consider the problem of learning undirected graphical models on trees from corrupted data. Recently Katiyar, Shah, and Caramanis showed that it is possible to recover trees from noisy binary data up to a small equivalence class of possible trees. Another paper by Katiyar, Hoffmann, and Caramanis follows a similar pattern for the Gaussian case. By framing this as a special phylogenetic recovery problem we largely generalize these two settings. Using the framework of linear latent tree models we discuss tree identifiability for binary data under a continuous corruption model (e.g. black/white images with greyscale corruption). For the Ising and the Gaussian tree model we also provide a characterisation of when the Chow-Liu algorithm consistently learns the underlying tree from the noisy data., Postprint (author's final draft)

L’arribada de les noves tecnologies ha tingut un gran impacte en el tenis, específicament en l’obtenció i anàlisi de les dades. Les noves eines d’enregistrament del moviment permeten la recopilació precisa de variables fisiològiques determinants les quals s’estudien per avaluar el rendiment i la càrrega física dels jugadors. Generalment es parla de la distància recorreguda (metres), percentatge del temps de joc efectiu, el nombre de partits disputats, el nombre de rally’s, entre d’altres. Amb les dades recopilades s’elaboren mètodes estadístics per identificar patrons sobre el rendiment dels jugadors i aquesta informació ajuda als entrenadors i jugadors a ajustar les estratègies de joc i a elaborar programes d’entrenament específics. L’objectiu d’aquest estudi és aprendre sobre les variables de càrrega física del tennis mitjançant dades reals de partits jugats als tornejos dels Grand Slams, per dues competicions ATP i WTA analitzades de forma independent, utilitzant el mètode de clustering K-means. Les dades s’han obtingut d’un domini públic, GitHub, i consisteixen en conjunts de dades on es registra informació punt per punt dels 4 tornejos de Grand Slam (Open Australia, Roland Garros, Open EEUU i Wimbledon), tant per la competició ATP com WTA, des del 2016 fins el 2021. El resultat de l’estudi actual ha estat classificar els partits i els jugadors segons les característiques que aquests presenten. S’han distingit 4 tipus de partits diferents tant per la competició masculina (ATP) com per la femenina (WTA) i tres tipus de tennistes diferents, per ambdós sexes, cada un amb particularitats específiques que els defineixen.

L’estadística esportiva és un àmbit que ha agafat gran importància els últims anys amb la disponibilitat de diferents recursos com articles científics i llibres en revistes especialitzades, la divulgació a partir de conferències, podcast i pel·lícules de l’àmbit que es poden trobar per exemple a la secció d’estadística esportiva de l’ American S tatistics Association (ASA) ASA). Tot i que hi ha estadístics o acadèmics que contribueixen en aquest àmbit, no es disposa ni es coneix encara d’un cens d’aquests professionals. És per això que l’ objectiu d’aquest TFG ha estat la creació d’un cens de tots els i nvestigadors i acadèmics especialitzats en estadística esportiva . Aquest primer cens consta de 191 investigadors de 26 països d’arreu del món dels quals se’ls hi ha fet arribar una enquesta que ha permès conèixer i ampliar el perfil d’aquests . La taxa de r esposta de l’enquesta ha estat del 50,49% (n = 103). Del total dels 103 investigadors, se n’han seleccionat 85 i s’ha pogut descriure el perfil d’aquests (tres cl u sters ) a partir de les tècniques de clustering k means i KAMILA. El primer cl u ster està forma t pels pioners o els que han impulsat per primera vegada l’estadística esportiva en el seu propi país (ex: USA, UK, Itàlia) i ja són acadèmics molt sèniors , en el segon cl u ster hi ha un grup majoritari d’edats entre 18 29 i 30 44 anys i són els que han continuat el que van crear els del primer cl u ster i a més hi ha la nova fornada d’investigadors més joves que combinen tant l’acadèmia com la indústria de l’esport amb experiència com a nalista en alguna organització o equip professional , mentre que al tercer cl u ster hi ha estadístics australians reconeguts sobretot en el camp de Sports Science i Exercise Medicine El cens realitzat ha ajudat a crear un primer mapa mundial i veure on es t roben distribuïts tots aquest s investigadors . Sembla c onfirma r se que Estats Units és el país líder en aquest àmbit amb un 24,71% dels investigadors i el segueix

L’objectiu principal del treball consisteix en la creació d’un indicador sintètic que resumeixi la informació continguda en diferents indicadors parcials. Aquest indicador servirà per observar el posicionament dels països Europeus pel que fa al seu benestar.

This paper provides a simulation and optimisation-based system to combine public transport (PT) with ride-pooling services (RP). According to the International Transport Forum (ITF), the RP could be established as a feeder of PT and included as the first or last leg of the journey with the option of transferring to/from PT in between. The system contains a dispatching core that uses an optimisation model with heuristic parameters to quickly analyse the potential permutations for each request. This topic is frequently based on simplistic modelling in the literature, and it has not been extensively tested in major urban regions. The whole metropolitan region of Barcelona is employed in this study, with a large realistic simulation model encompassing a 20 × 15 km area with a PT network of about 3000 stations and 300 route lines and nearly 114,000 traffic links. This enables for a more accurate evaluation of system performance and trip quality computation., This research was funded by an Industrial PhD by the Generalitat de Catalunya and the company PTV-AG under Grant DI-071 2019; and a research project of the Spanish R+D Programs under Grant PID2020-112967GB-C31., Peer Reviewed, Postprint (published version)

Les entitats financeres de crèdit necessiten obtenir previsions precises de les futures taxes d'impagament, ja que són clau per a la determinació d'estratègies adequades de gestió del risc. Unes bones previsions (és a dir, baixes taxes d'impagament) permeten a les empreses financeres ampliar la seva cartera. Per contra, la identificació d'escenaris a la baixa (és a dir, altes taxes d'impagament) condueix a una restricció de les polítiques de concessió i a augments dels fons de capital per cobrir possibles pèrdues. Atesa la relació estreta entre els impagaments en una societat i l'estat de l'economia, s'ha demostrat que les variables macroeconòmiques són forts predictors de la taxa d'impagament. Com a conseqüència de l'estructura temporal de les dades i la disponibilitat de diverses variables, s'han considerat tècniques de sèries temporals multivariants. Aquest projecte es centra en un d'aquests models: el model ARDL i l'UECM. S'han seguit dos enfocaments: el primer intenta construir un model únic, mentre que el segon construeix diversos models bivariants simples i afegeix les seves prediccions. L'últim d'aquests enfocaments ha demostrat ser millor, fins i tot superant el model de referència univariant en alguns casos., Las entidades financieras de crédito necesitan obtener previsiones precisas de sus futuras tasas de impago, ya que éstas son clave para la determinación de estrategias adecuadas de gestión del riesgo. Unas buenas previsiones (es decir, bajas tasas de impago) permiten a las empresas financieras ampliar su cartera. Por el contrario, la identificación de escenarios bajistas (es decir, altas tasas de impago) conduce a una restricción de las políticas de concesión y a aumentos de los fondos de capital para cubrir posibles pérdidas. Dada la estrecha relación entre los impagos en una sociedad y el estado de la economía, se ha demostrado que las variables macroeconómicas son fuertes predictores de la tasa de impago. Como consecuencia de la estructura temporal de los datos y la disponibilidad de distintas variables, se han considerado técnicas de series temporales multivariantes. Este proyecto se centra en uno de estos modelos: el modelo ARDL y el UECM. Se han seguido dos enfoques: el primero trata de construir un modelo único, mientras que el segundo construye varios modelos bivariantes simples y agrega sus predicciones. El último de estos enfoques ha demostrado ser mejor, superando incluso al modelo de referencia univariante en algunos casos., Credit financial entities need to obtain accurate forecasts of their future default rates as these are key for the determination of adequate risk management strategies. Good previsions (i.e. low default rates) allow financial companies to expand their portfolio. In contrast, the identification of downside/stream scenarios (i.e. high default rates) leads to a restriction of concession policies and increases in capital funds to cover potential losses. Given the close relationship between defaults in a society and the state of the economy, macroeconomic variables have been demonstrated to be strong predictors of the default rate. Given the time structure of the data and the availability of different variables, multivariate time series techniques have been considered. This project focuses on one of these models: the ARDL model and the UECM. Two approaches have been followed: the first approach tries to construct a single model while the second approach builds various simple bivariate models and aggregates their predictions. The latter of these approaches has been proven to be better, even outperforming the univariate benchmark model in some cases.

The mental health landscape worldwide has been profoundly affected by the COVID-19 pandemic, with an escalating rise in stress, anxiety, and depression. As a result, the urgency to create and apply efficacious methods for averting and addressing these mental health issues has never been greater. In this context COVICAT is a research project that is being carried out in Catalonia, Spain. The project is led by the Germans Trias i Pujol Research Institute (IGTP) and the Barcelona Institute for Global Health (ISGlobal). The goal of COVICAT is to evaluate the impact of the COVID-19 pandemic on the Catalan population. This master's thesis aims to leverage machine learning methodologies for psychological risk assessment using the COVICAT dataset (N = 9,398), contributing to healthcare practices by facilitating the initial classification of patients requiring psychological intervention. The study focuses on predicting four outcome variables reflecting patients' psychological wellbeing, one of which encapsulates a patient's self-perceived mental state. Initial exploratory data analysis and data engineering were performed to prepare the dataset for machine learning applications, including bagging, boosting, and stacking models. Each model underwent an analogous pipeline with minor variations, dictated by the distinct nature of the respective model. Post-training, these models were evaluated on unseen data, generating several performance metrics. The F1-score, designated as the critical criterion by the subject matter experts, was employed to gauge the models' efficacy. Based on these evaluations, the optimal predictive model was selected for each target. The resulting models delivered performance surpassing the adequate standard established initially by the experts (F1 = 0.6). These models aim to augment, not supplant, professional healthcare diagnoses by enabling efficient patient triage for psychological assistance.

The single-index model is a non-parametric model that is popular in statistics for the estimation of risk assessment, thanks to its flexibility. Many studies have approached the problem of finding the best estimators for the model parameters and smooth function. However, most of the real-world data is asymmetric. In this thesis, we consider the application of a Box-Cox transformation on the response variable to improve the performances of the single-index model. Indeed, the Box-Cox transformation addresses the skewness of the response and transform it into a normally-distributed variables. We consider also the log transformation of the response variable. We show through a simulation study that the parameters of the models fitted with the transformed response have more similarity with the true parameters, which shows that the model is more reliable. Finally, we perform a study on a real-world data set from the car insurance industry, where we build a single-index model to predict the expected cost per claim based on a set of variables, that concern the demographic characteristic and the driving style of the customer. We assess the performance of the model fitted on the original data as well the models that use the log and the Box-Cox transformation, that obtain better results than the original one.

This paper shows the added value of using the existing specific domain knowledge to generate new derivated variables to complement a target dataset and the benefits of including these new variables into further data analysis methods. The main contribution of the paper is to propose a methodology to generate these new variables as a part of preprocessing, under a double approach: creating 2nd generation know dge-driven variables, catching the experts criteria used for reasoning on the field or 3rd generation data-driven indicators, these created by clustering original variables. And Data Mining and Artificial Intelligence techniques like Clustering or Traffic light Panels help to obtain successful results. Some results of the project INSESS-COVID19 are presented, Basic descriptive analysis gives simple results that eventhough they are useful to support basic policy-making, especially in health, a much richer global perspective is acquired after including derivated variables. When 2nd generation variables are available and can be introduced in the method for creating 3rd generation data, added value is obtained from both basic analysis and building new data-driven indicators., Peer Reviewed, Postprint (author's final draft)

The analysis of time series studies linking daily counts of a health indicator with environmental variables (e.g., mortality or hospital admissions with air pollution concentrations or temperature; or motor vehicle crashes with temperature) is usually conducted with Poisson regression models controlling for long-term and seasonal trends using temporal strata. When the study includes multiple zones, analysts usually apply a two-stage approach: first, each zone is analyzed separately, and the resulting zone-specific estimates are then combined using meta-analysis. This approach allows zone-specific control for trends. A one-stage approach uses spatio-temporal strata and could be seen as a particular case of the case–time series framework recently proposed. However, the number of strata can escalate very rapidly in a long time series with many zones. A computationally efficient alternative is to fit a conditional Poisson regression model, avoiding the estimation of the nuisance strata. To allow for zone-specific effects, we propose a conditional Poisson regression model with a random slope, although available frequentist software does not implement this model. Here, we implement our approach in the Bayesian paradigm, which also facilitates the inclusion of spatial patterns in the effect of interest. We also provide a possible extension to deal with overdispersed data. We first introduce the equations of the framework and then illustrate their application to data from a previously published study on the effects of temperature on the risk of motor vehicle crashes. We provide R code and a semi-synthetic dataset to reproduce all analyses presented., We acknowledge support from the grant CEX2018- 000806-S funded by MCIN/AEI/ 10.13039/501100011033, support from Ministry of Research and Universities of the Government of Catalonia (2021 SGR 01563) and support from the Generalitat de Catalunya through the CERCA Program., Peer Reviewed, Postprint (published version)

Public Bicycle Sharing Systems have spread in many cities for the last decade. The need of analysis tools to predict the behavior or estimate balancing needs has fostered a wide set of approaches that consider many variables. Often, these approaches use a single scenario to evaluate their algorithms, and little is known about the applicability of such algorithms in cities of different sizes. In this paper, we evaluate the performance of widely known prediction algorithms for three sized scenarios: a small system, with around 20 docking stations, a medium-sized one, with 400+ docking stations, and a large one, with more than 1500 stations. The results show that Prophet and Random Forest are the prediction algorithms with more consistent results, and that small systems often have not enough data for the algorithms to perform a solid work., Partially supported by PID2021-122136OB-C21 by Ministerio de Ciencia e Innovación/AEI: (10.13039/501100011033/FEDER, UE), and 2021SGR00613-(ADBD) by Generalitat de Cataluña. The authors would like to thank Logroño City Hall for providing the access to the data., Peer Reviewed, Postprint (published version)

Breast cancer stands as a prominent global health concern, consistently ranking among the most prevalent types of cancer affecting women. In this work, the focus is put on invasive ductal carcinoma and invasive lobular carcinoma, which are two most common types of breast cancer. Transfer learning techniques are applied to implement a Convolutional Neural Network (CNN) architecture that accurately classifies histological images accurately based on the subtype of cancer. Next, an integrative analysis is done with the artificial descriptors obtained from the CNN and gene expression RNA-Seq data to see the associations between the two datasets. Finally, a gene set enrichment analysis is performed with the aim of providing a biological interpretation to the artificial descriptors obtained from the CNN, by correlating them with biological pathways relevant to these two subtypes of breast cancer.

Aquest estudi consisteix en determinar els perfils de jugadors de futbol que van disputar la Champions League entre el 2009/2010 i el 2021/2022 per tal de poder entendre les característiques i aquells factors que podrien determinar el rendiment individual i influir en l’èxit col·lectiu. En els últims anys, l'ús de les dades i l'anàlisi estadístic s'ha convertit en una part integral del món de l'esport. Les empreses i els equips han utilitzat les dades per comprendre millor els patrons i les tendències que afecten el rendiment dels esportistes i dels equips. Això ha donat lloc a l'aparició dels sports analytics, una disciplina que s'utilitza per analitzar dades esportives i obtenir informació útil per als equips, els entrenadors i jugadors. La tecnologia ha estat un dels factors clau en l'evolució dels 'sports analytics i la soccer analytics. Les dades esportives es recopilen ara de manera més precisa i detallada que mai abans, gràcies a les tecnologies com les càmeres de seguiment, els sensors i les xarxes de sensors. Això ha permès als equips obtenir dades més precises sobre el rendiment dels jugadors i sobre les condicions del joc. La creació d'una base de dades des de zero, la búsqueda de dades fiables i recolzades per la comunitat, l’estructuració de la base de dades i l’elecció dels factors claus de rendiment són, molt importants per poder obtenir resultats precisos i que s'adeqüin als objectius plantejats són els pilars fonamentals d’aquest estudi. S’han pogut determinar diversos perfils de jugadors en funció de les característiques seleccionades, com són les defensives, les ofensives, i les de construcció de joc. S’observen clares diferències entre els perfils de jugadors segons les característiques defensives proposades, mentres que les diferències entre els jugadors que formen part dels clústers de construcció de joc no són tan significatives. De la mateixa manera, els resultats obtinguts es poden interpretar de diverses maneres en funció de l’objectiu que

En aquesta tesi s’ha desenvolupat un algoritme anomenat ”Epidemiological Foci Relating Infections by Distance”(EpiFRIenDs) orientat a la detecció punts calents (hotspots) de malària a partir d’unes dades geolocalitzades de casos detectats de malària. L’algoritme està basat en l’algoritme Friends of friends (FoF) utilitzat generalment en cosmologia per detectar estructures cosmològiques com halos de matèria fosca (component hipotètic d’una galàxia que embolcalla el disc galàctic i s’estén més enllà de l’extrem visible de la galàxia). El mètode consistirà en relacionar tota una xarxa de punts entre ells a partir de la cerca de tots els punts més propers a una distància que s’anomenarà linking distance. Es farà servir l’algoritme KDTrees que permetrà baixar notablement el cost computacional per xarxes de punts no excessivament grans. Del resultat generat es cercaran totes les relacions entre les agrupacions a partir de punts coincidents de forma que tots els punts entre els quals existeixi un camí de punts amb distàncies menors a la linking distance quedaran agrupats en una sola agrupació que s’anomenarà cluster. D’aquests clusters s’extrauran dades d’interès epidemiològic com la ràtio de casos positius. L’algoritme presentarà un enfocament innovador en comparació a mètodes anteriors, ja que es centrarà a trobar clusters en la seva forma exacta (i no una zona aproximada) d’una forma computacionalment eficaç, permetent d’aquesta forma treballar amb poblacions que puguin presentar distribucions poc convencionals com per exemple a la vora d’un riu i de forma personalitzable. Addicionalment, es proposarà una optimització alguns paràmetres seleccionables amb l’objectiu de trobar els clusters estadísticament més significants possibles., EpiFRIenDs (Epidemiological Foci Relating Infections by Distance) es un algoritmo novedoso basado en el Friends of Friends (foF) que se ha desarrollado para detectar zonas de otra transmisión de malaria. En esta tesis se explicará su funcionamiento, opciones de optimización y uso., EpiFRIenDs (Epidemiological Foci Relating Infections by Distance) is a new algorithm based on Friends of Friends (foF) that has been developed to detect areas of other malaria transmission. This thesis will explain its operation, optimization options and its use.

Aquest treball de recerca representa una oportunitat per aprofundir en l'anàlisi espai-temporal de l'ús del Bicing a Barcelona, utilitzant metodologies estadístiques avançades i eines com el Shiny del llenguatge de programació R. Mitjançant l'estudi d'aquestes dades, s'aspira a comprendre millor els patrons d'ús de les bicicletes i identificar àrees d'optimització de les estacions en el sistema del Bicing. Aquest enfocament pot contribuir a la millora del transport urbà sostenible i promoure una mobilitat més saludable i ecològica a la ciutat.

Measurement of external load in players provides objective information to optimise the weekly balance between training and recovery to improve performance and prevent injuries. Our aim was to evaluate the incidence of sports-related muscle injuries of the lower limb in relation to external load, measured by global positioning system (GPS), in football players. A descriptive study was carried out. Data were collected from 71 professional male football players (30 professionals and 41 youth players) from an elite football club competing in the Spanish and European League in the 2017–2018 season. As external load variables, we measured High Metabolic Load Distance (HMLD), High Speed Running (HSR), Player Load (PL), and Total Distance (TD) through GPS. Injury rate (IR) was calculated both in relation to such GPS load metrics and to load exposure time. We considered categories (youth and professional), playing positions (centre back, full back, midfielder, and forward), and training day with respect to match-day (-4MD, -3MD, -2MD, -1MD, MD, +1MD, +2MD). The GPS load metrics HMLD, HSR, PL, and TD showed very similar patterns across categories and positions, but varied according to training session or MD. The highest loads were observed on MD and three days prior to the match (-3MD). Similarly, the overall IR, both calculated per load exposure time and per GPS load metrics, was highest on MD and -3MD. Again, no differences were observed between youth and professional players. Midfielders demonstrated the highest IR in all metrics, followed by the forwards. In conclusion, this study suggests that external load and incidence of muscle injuries are directly proportional. Therefore, the measurement of more external load variables other than load exposure time, such as the GPS metrics HMLD, HSR, PL, and TD may help to describe the pattern and magnitude of injuries. Future studies based on ours may help to further improve the understanding of the incidence of muscle injuries on, JC and MC acknowledge the financial support from Ministerio de Ciencia e Innovación y Ministerio de Universidades (Ref: PID2019-104830RB-I00 / AEI DOI: 10.13039/501100011033)., Peer Reviewed, Postprint (published version)

En la actualidad, la temática de la salud mental está adquiriendo mayor relevancia, pero aún se tienen pocos registros y estudios al respecto. A través del presente trabajo, que analiza estadísticamente la dinámica natural del número de suicidios en España, se quiere contribuir a dar un enfoque científico a la temática y también a que la ideación suicida tenga la importancia y visibilidad que merece. El objetivo estadístico de este estudio es analizar, con fines predictivos, la evolución de los datos de suicidios mensuales en España a lo largo del tiempo entre 1990 y 2019. Para el análisis estadístico, se han obtenido registros de suicidios del Instituto Nacional de Estadística (INE) y se ha recurrido al análisis de series temporales. Concretamente, se ha aplicado la metodología ARIMA (autoregressive integrated moving average) clásica de Box-Jenkins y su extensión conocida como ARIMAX. El procedimiento de construcción de un modelo ARIMAX, ha permitido seleccionar al final un modelo con buena calidad de ajuste y capacidad predictiva. Además, los resultados obtenidos permiten predecir con cierta confianza el comportamiento que podría seguir el número de personas que deciden quitarse la vida si no se implantan medidas de prevención efectivas.

The purpose of this study is to determine Girona Cathedral’s thermal performance and to verify that it is suitable for containing historical pieces of art. We present the results of our analysis of the indoor thermal conditions during the period from January 2011 to December 2016. Real data were collected from strategically located probes inside the cathedral and from an outside weather station. The results were compared with an EnergyPlus computer model to verify the calculations and improve decision making. The model considers the envelope’s thermal inertia, the characteristics of the enclosure, and the passive systems for performing thermal analysis. The cathedral’s indoor environment is very stable. Because of a high capability of thermal-energy accumulation, a one-month thermal lag between indoor and outdoor temperatures exists. Furthermore, the maximum and minimum temperatures are dampened, with a two-degree variation above or below the outdoor average, depending on the season. The outdoor humidity can vary from 40% to 100% daily, whereas the indoor humidity variation is around 5%. This indoor stable condition protects the artistic objects in the building. This six-year monitoring and analysis allows for the determination that high-inertia buildings are suitable for displaying and preserving pieces of art and antique furniture, protecting them from deterioration., Peer Reviewed, Postprint (published version)

Citizens take part in different activities to satisfy their needs, to invest in their socio-economic progress, participate in social and health activities that improve their well-being. However, activity participation is influenced by many factors in the built environment, but also individual’s attributes. Herein we analyze activity participation and travel through sequence analysis. This method explores sequences of daily activity and travel employing techniques from the sequencing of events in the life course of individuals. Studying sequences of daily episodes (each activity and each trip) considers the entire trajectory of a person’s activity during a day while at the same time considering the number of activities, order of activities in a day, and their durations jointly. We applied this method to a sample of residents in the Metropolitan Area of Barcelona (RMB) in the 2018, 2019 and 2020 EMEF Travel Surveys. The EMEF2020 deserves a particular analysis since activity patterns are expected to vary compared to pre-COVID19 spread. We have focused on that fragmentation in activity participation over the mean among persons in specific gender, age, activity and transportation mode., This research was funded by PID2020-112967GB-C31 Spanish R+D Programs and by Secretaria d’Universitats-i-Recerca-Generalitat de Catalunya- 2017-SGR- 1749. The datasets were kindly shared by the Autoritat del Transport Metropolità (ATM). Their contribution to our research is gratefully acknowledged., Peer Reviewed, Postprint (published version)

El objetivo de este trabajo es estudiar la asociación entre las medidas antropométricas y la actividad física, y el riesgo de desarrollar un cáncer de cuello uterino o uno de sus precursores. A partir de la base de datos del estudio EPIC (European Prospective Investigation into Cancer and Nutrition) se realizarán varios tipos de análisis: modelos de riesgo proporcional de Cox para estimar hazard ratio y modelos de regresión logística condicional para estimar odds ratios. Además, se validarán ambos tipos de modelos, en concreto, el de regresión logística condicional mediante un ajuste de bondad específico para datos emparejados.

During the current emergency of Pandemic disease of Covid-19, the importance of online learning strategy has much increased due to closure of schools, colleges and universities in most part of the globe. To evaluate the effectiveness of online learning, a statistical study has been conducted on the academic achievements of postgraduate students learned by online versus face-to-face learning modes. The statistical analysis of academic achievements of postgraduate students has been done by SPSS-20 statistical tool. The t-test analysis showed that the t-value was 0.549 and p-value was 0.583. The significance level for t-test analysis was 0.05. As, the p-value was more that 0.05 (p˃0.05) so, no significant relation has been found in the academic achievements of the students learned by online and face to face modes of learning. The null hypothesis (Ho) was proved true and the alternative hypothesis (H1) rejected. The reasons for no significant relation were abruptly shifting of learning modes from face-to-face to online, slow working of LMS accounts due to over burden, login and voice problems during the online classes. Overall, the students performed better in the online modes of learning. Some students performed good in face-to-face learning and performed poor in online modes of learning. So, to generalize, it is necessary to do more research on online modes of learning under regional context, Peer Reviewed

Objective Direct cortical stimulation (DCS) is standard for intracranial presurgical evaluation in drug-resistant epilepsy (DRE). Few studies have reported levels of concordance between spontaneous seizure generators and triggered seizures during DCS. The present work reports validity measures of DCS for detecting the seizure onset zone (SOZ) during stereoelectroencephalography (SEEG). Methods We evaluated all patients who underwent SEEG evaluation at our epilepsy center between 2013 and 2019. Data were analyzed using contingency tables. Validity measures of the diagnostic test were computed for all patients evaluated with DCS and for seizure free patients. Results Fifty-eight consecutive patients were evaluated through DCS. One hundred seventy-three clinical seizures were elicited with DCS. Electroclinical identical to spontaneous seizures were considered true positive (TP) seizures. They showed a high specificity (96.9%) for detecting the SOZ in patients that remained seizure free one year after treatment. Sensitivity was low (23.0%), and a high percentage of false-negative stimulations was documented in the SOZ. The accuracy was 87.9%. Conclusions DCS is a technique with high specificity but a low sensitivity for the localization of the SOZ. The DCS validity measures need to be known when considered for surgical decisions. The interpretation of DCS-triggered seizures and the differentiation of true-positive vs false-positive seizures should be carefully evaluated. Significance DCS seizure triggering is highly specific for SOZ localization., This work received support from the project “Clúster Emergent del Cervell Humà” (CECH) and the European Regional Development Fund under the framework of the ERFD Operative Programme for Catalonia., Peer Reviewed, Postprint (author's final draft)