Pasar al contenido principal

Resultados relacionados

Resultados totales (Incluyendo duplicados): 16
Encontrada(s) 1 página(s)

RiuNet. Repositorio Institucional de la Universitat Politécnica de Valéncia
Publicación Artículo científico (JournalArticle)

Mapping the scientific structure of organization and management of enterprises using complex networks

  • Olivares-Gil, Alicia
  • Arnaiz-Rodríguez, Adrián
  • Ramírez-Sanz, José Miguel
  • Garrido-Labrador, José Luis
  • Ahedo, Virginia
  • García-Osorio, César
  • Santos, José Ignacio
  • Galán, José Manuel
[EN] Understanding the scientific and social structure of a discipline is a fundamental aspect for scientific evaluation processes, identifying trends and niches, and balancing the trade-off between exploitation and exploration in research. In the present contribution, the production of doctoral theses is used as a proxy to analyze the scientific structure of the knowledge area of business organization in Spain. To that end, a complex networks approach is selected, and two different networks are built: (i) the social network of co-participation in thesis examining committees and thesis supervision, and (ii) a bipartite network of theses and thesis descriptors. The former has a modular structure that is partially explained by thematic specialization in different subdisciplines. The latter serves to assess the interdisciplinary structure of the discipline, as it enables the characterization of affinity levels between fields, research poles and thematic clusters. Our results have implications for the scientific evaluation and formal definition of related fields., The authors acknowledge financial support from the and Spanish Ministry of Science, Innovation Universities (RED2018-102518-T), the Spanish State Research Agency (PID2020118906GB-I00 and PID2020-119894GB-I00 via AEI/10.13039/501100011033), the Junta de Castilla y León – Consejería de Educación (BU055P20), Fundación La Caixa (2020/00062/001) and from NVIDIA Corporation and its donation of the TITAN Xp GPUs that facilitated this research. This work was partially supported by the European Social Fund, as the authors José Miguel Ramírez-Sanz, José Luis Garrido-Labrador and Alicia OlivaresGil are the recipient of a predoctoral grant from the Department of Education of Junta de Castilla y León (VA) (ORDEN EDU/875/2021). In addition, this work was also partially supported by the Generalitat Valenciana via its Conselleria de Innovación, Universidades, Ciencia y Sociedad Digital, as Adrián Arnaiz is recipicient of a predoctoral grant. The authors would also like to thank Dr. Manzanedo, Dra. Saiz-Bárcena, Dr. Solé Parellada, Dr. Izquierdo and Dr. del Olmo for their insightful help to improve the manuscript.
Repositorio Institucional de la Universidad de Burgos (RIUBU)
Publicación Artículo científico (JournalArticle)

Experimental evaluation of ensemble classifiers for imbalance in Big Data

  • Juez Gil, Mario
  • Arnaiz González, Álvar
  • Rodríguez Diez, Juan José
  • García Osorio, César
Datasets are growing in size and complexity at a pace never seen before, forming ever larger datasets known as Big Data. A common problem for classification, especially in Big Data, is that the numerous examples of the different classes might not be balanced. Some decades ago, imbalanced classification was therefore introduced, to correct the tendency of classifiers that show bias in favor of the majority class and that ignore the minority one. To date, although the number of imbalanced classification methods have increased, they continue to focus on normal-sized datasets and not on the new reality of Big Data. In this paper, in-depth experimentation with ensemble classifiers is conducted in the context of imbalanced Big Data classification, using two popular ensemble families (Bagging and Boosting) and different resampling methods. All the experimentation was launched in Spark clusters, comparing ensemble performance and execution times with statistical test results, including the newest ones based on the Bayesian approach. One very interesting conclusion from the study was that simpler methods applied to unbalanced datasets in the context of Big Data provided better results than complex methods. The additional complexity of some of the sophisticated methods, which appear necessary to process and to reduce imbalance in normal-sized datasets were not effective for imbalanced Big Data., “la Caixa” Foundation, Spain, under agreement LCF/PR/PR18/51130007. This work was supported by the Junta de Castilla y León, Spain under project BU055P20 (JCyL/FEDER, UE) co-financed through European Union FEDER funds, and by the Consejería de Educación of the Junta de Castilla y León and the European Social Fund, Spain through a pre-doctoral grant (EDU/1100/2017).
Repositorio Institucional de la Universidad de Burgos (RIUBU)
Publicación Artículo científico (JournalArticle)

When is resampling beneficial for feature selection with imbalanced wide data?

  • Ramos Pérez, Ismael
  • Arnaiz González, Álvar
  • Rodríguez Diez, Juan José
  • García Osorio, César
This paper studies the effects that combinations of balancing and feature selection techniques have on wide data (many more attributes than instances) when different classifiers are used. For this, an extensive study is done using 14 datasets, 3 balancing strategies, and 7 feature selection algorithms. The evaluation is carried out using 5 classification algorithms, analyzing the results for different percentages of selected features, and establishing the statistical significance using Bayesian tests. Some general conclusions of the study are that it is better to use RUS before the feature selection, while ROS and SMOTE offer better results when applied afterwards. Additionally, specific results are also obtained depending on the classifier used, for example, for Gaussian SVM the best performance is obtained when the feature selection is done with SVM-RFE before balancing the data with RUS., “La Caixa” Foundation, under agreement LCF/PR/PR18/51130007. This work was also supported by the Junta de Castilla León under project BU055P20 (JCyL/FEDER, UE) and by the Ministry of Science and Innovation under project PID2020-119894GB-I00, co-financed through European Union FEDER funds .
Repositorio Institucional de la Universidad de Burgos (RIUBU)
Publicación Artículo científico (JournalArticle)

Approx-SMOTE: Fast SMOTE for Big Data on Apache Spark

  • Juez Gil, Mario
  • Arnaiz González, Álvar
  • Rodríguez Diez, Juan José
  • López Nozal, Carlos
  • García Osorio, César
One of the main goals of Big Data research, is to find new data mining methods that are able to process large amounts of data in acceptable times. In Big Data classification, as in traditional classification, class imbalance is a common problem that must be addressed, in the case of Big Data also looking for a solution that can be applied in an acceptable execution time. In this paper we present Approx-SMOTE, a parallel implementation of the SMOTE algorithm for the Apache Spark framework. The key difference with the original SMOTE, besides parallelism, is that it uses an approximated version of k-Nearest Neighbor which makes it highly scalable. Although an implementation of SMOTE for Big Data already exists (SMOTE-BD), it uses an exact Nearest Neighbor search, which does not make it entirely scalable. Approx-SMOTE on the other hand is able to achieve up to 30 times faster run times without sacrificing the improved classification performance offered by the original SMOTE., “La Caixa” Foundation, under agreement LCF/PR/PR18/51130007. This work was supported by the Junta de Castilla y León under project BU055P20 and by the Ministry of Science and Innovation of Spain under project PID2020-119894 GB-I00, co-financed through European Union FEDER funds. It also was supported through Consejería de Educación of the Junta de Castilla y León and the European Social Fund through a pre-doctoral grant (EDU/1100/2017). This material is based upon work supported by Google Cloud.
Repositorio Institucional de la Universidad de Burgos (RIUBU)
Publicación Artículo científico (JournalArticle)

Rotation Forest for Big Data

  • Juez Gil, Mario
  • Arnaiz González, Álvar
  • Rodríguez Diez, Juan José
  • López Nozal, Carlos
  • García Osorio, César
The Rotation Forest classifier is a successful ensemble method for a wide variety of data mining applications. However, the way in which Rotation Forest transforms the feature space through PCA, although powerful, penalizes training and prediction times, making it unfeasible for Big Data. In this paper, a MapReduce Rotation Forest and its implementation under the Spark framework are presented. The proposed MapReduce Rotation Forest behaves in the same way as the standard Rotation Forest, training the base classifiers on a rotated space, but using a functional implementation of the rotation that enables its execution in Big Data frameworks. Experimental results are obtained using different cloud-based cluster configurations. Bayesian tests are used to validate the method against two ensembles for Big Data: Random Forest and PCARDE classifiers. Our proposal incorporates the parallelization of both the PCA calculation and the tree training, providing a scalable solution that retains the performance of the original Rotation Forest and achieves a competitive execution time (in average, at training, more than 3 times faster than other PCA-based alternatives). In addition, extensive experimentation shows that by setting some parameters of the classifier (i.e., bootstrap sample size, number of trees, and number of rotations), the execution time is reduced with no significant loss of performance using a small ensemble., This work was supported through project TIN2015-67534-P (MINECO/FEDER, UE) of the Ministerio de Economía y Competitividad of the Spanish Government, projects BU085P17 and BU055P20 (JCyL/FEDER, UE) of the Junta de Castilla y León, Spain (both projects co-financed through European Union FEDER funds), and by the Consejería de Educación of the Junta de Castilla y León, Spain and the European Social Fund through a pre-doctoral grant (EDU/1100/2017). The project leading to these results has received also funding from “la Caixa” Foundation, Spain , under agreement LCF/PR/PR18/51130007. This material is based upon work supported by Google Cloud.
Repositorio Institucional de la Universidad de Burgos (RIUBU)
Publicación Artículo científico (JournalArticle)

Mapping the scientific structure of organization and management of enterprises using complex networks

  • Olivares Gil, Alicia
  • Arnaiz Rodríguez, Adrián
  • Ramírez Sanz, José Miguel
  • Garrido Labrador, José Luis
  • Ahedo García, Virginia
  • García Osorio, César
  • Santos Martín, José Ignacio
  • Galán Ordax, José Manuel
Understanding the scientific and social structure of a discipline is a fundamental aspect for scientific evaluation processes, identifying trends and niches, and balancing the trade-off between exploitation and exploration in research. In the present contribution, the production of doctoral theses is used as a proxy to analyze the scientific structure of the knowledge area of business organization in Spain. To that end, a complex networks approach is selected, and two different networks are built: (i) the social network of co-participation in thesis examining committees and thesis supervision, and (ii) a bipartite network of theses and thesis descriptors. The former has a modular structure that is partially explained by thematic specialization in different subdisciplines. The latter serves to assess the interdisciplinary structure of the discipline, as it enables the characterization of affinity levels between fields, research poles and thematic clusters. Our results have implications for the scientific evaluation and formal definition of related fields., Spanish Ministry of Science, Innovation and Universities (RED2018-102518-T), the Spanish State Research Agency (PID2020-118906GB-I00 and PID2020-119894GB-I00 via AEI/10.13039/501100011033), the Junta de Castilla y León – Consejería de Educación (BU055P20), Fundación La Caixa (2020/00062/001) and from NVIDIA Corporation and its donation of the TITAN Xp GPUs that facilitated this research. This work was partially supported by the European Social Fund, as the authors José Miguel Ramírez-Sanz, José Luis Garrido-Labrador and Alicia Olivares-Gil are the recipient of a predoctoral grant from the Department of Education of Junta de Castilla y León (VA) (ORDEN EDU/875/2021). In addition, this work was also partially supported by the Generalitat Valenciana via its Conselleria de Innovación, Universidades, Ciencia y Sociedad Digital, as Adrián Arnaiz is recipicient of a predoctoral grant.
Repositorio Institucional de la Universidad de Burgos (RIUBU)
Publicación Artículo científico (JournalArticle)

When is resampling beneficial for feature selection with imbalanced wide data?

  • Ramos Pérez, Ismael
  • Arnaiz González, Álvar
  • Rodríguez Diez, Juan José
  • García Osorio, César
This paper studies the effects that combinations of balancing and feature selection techniques have on wide data (many more attributes than instances) when different classifiers are used. For this, an extensive study is done using 14 datasets, 3 balancing strategies, and 7 feature selection algorithms. The evaluation is carried out using 5 classification algorithms, analyzing the results for different percentages of selected features, and establishing the statistical significance using Bayesian tests. Some general conclusions of the study are that it is better to use RUS before the feature selection, while ROS and SMOTE offer better results when applied afterwards. Additionally, specific results are also obtained depending on the classifier used, for example, for Gaussian SVM the best performance is obtained when the feature selection is done with SVM-RFE before balancing the data with RUS., The project leading to these results has received funding from “la Caixa” Foundation, under agreement LCF/PR/PR18/51130007. This work was also supported by the Junta de Castilla León under project BU055P20 (JCyL/FEDER, UE) and by the Ministry of Science and Innovation under project PID2020-119894GB-I00, co-financed through European Union FEDER funds.
Repositorio Institucional de la Universidad de Burgos (RIUBU)
Publicación Artículo científico (JournalArticle)

Experimental Assessment of Feature Extraction Techniques Applied to the Identification of Properties of Common Objects, Using a Radar System

  • Diez Pastor, José Francisco
  • Latorre Carmona, Pedro
  • Garrido Labrador, José Luis
  • Ramírez Sanz, José Miguel
  • Rodríguez Diez, Juan José
Radar technology has evolved considerably in the last few decades. There are many areas where radar systems are applied, including air traffic control in airports, ocean surveillance, and research systems, to cite a few. Other types of sensors have recently appeared, which allow tracking sub-millimeter motion with high speed and accuracy rates. These millimeter-wave radars are giving rise to myriad new applications, from the recognition of the material close objects are made, to the recognition of hand gestures. They have also been recently used to identify how a person interacts with digital devices through the physical environment (Tangible User Interfaces, TUIs). In this case, the radar is used to detect the orientation, movement, or distance from the objects to the user’s hands or the digital device. This paper presents a thoughtful comparative analysis of different feature extraction techniques and classification strategies applied on a series of datasets that cover problems such as the identification of materials, element counting, or determining the orientation and distance of objects to the sensor. The results outperform previous works using these datasets, especially when the accuracy was lowest, showing the benefits feature extraction techniques have on classification performance., This work was supported by the Spanish Ministry of Science and Innovation under project PID2020-119894GB-I00, Junta de Castilla y León under project BU055P20 (JCyL/FEDER, UE) co-financed through European Union FEDER funds. José Luis Garrido-Labrador was supported by the predoctoral grant (BDNS 510149) awarded by the Universidad de Burgos, Spain.
Repositorio Institucional de la Universidad de Burgos (RIUBU)
Publicación Artículo científico (JournalArticle)

An experiment on animal re-identification from video

  • Kuncheva, Ludmila I. .
  • Garrido Labrador, José Luis
  • Ramos Pérez, Ismael
  • Hennessey, Samuel L.
  • Rodríguez Diez, Juan José
In the face of the global concern about climate change and endangered ecosystems, monitoring individual animals is of paramount importance. Computer vision methods for animal recognition and re-identification from video or image collections are a modern alternative to more traditional but intrusive methods such as tagging or branding. While there are many studies reporting results on various animal re-identification databases, there is a notable lack of comparative studies between different classification methods. In this paper we offer a comparison of 25 classification methods including linear, non-linear and ensemble models, as well as deep learning networks. Since the animal databases are vastly different in characteristics and difficulty, we propose an experimental protocol that can be applied to a chosen data collections. We use a publicly available database of five video clips, each containing multiple identities (9 to 27), where the animals are typically present as a group in each video frame. Our experiment involves five data representations: colour, shape, texture, and two feature spaces extracted by deep learning. In our experiments, simpler models (linear classifiers) and just colour feature space gave the best classification accuracy, demonstrating the importance of running a comparative study before resorting to complex, time-consuming, and potentially less robust methods., This work is supported by the UKRI Centre for Doctoral Training in Artificial Intelligence, Machine Learning and Advanced Computing (AIMLAC), funded by grant EP/S023992/1. This work is also supported by the Junta de Castilla León under project BU055P20 (JCyL/FEDER, UE), and the Ministry of Science and Innovation under project PID2020-119894 GB-I00 co-financed through European Union FEDER funds. J.L. Garrido-Labrador is supported through Consejería de Educación of the Junta de Castilla y León and the European Social Fund through a pre-doctoral grant (EDU/875/2021). I. Ramos-Perez is supported by the predoctoral grant (BDNS 510149) awarded by the Universidad de Burgos, Spain. J.J. Rodríguez was supported by mobility grant PRX21/00638 of the Spanish Ministry of Universities.
Repositorio Institucional de la Universidad de Burgos (RIUBU)
Publicación Artículo científico (JournalArticle)

Network analysis of co-participation in thesis examination committees in an academic field in Spain, Análisis de la red de co-participaciones en tribunales de tesis de un área científica en España

  • Garrido Labrador, José Luis
  • Ramírez Sanz, José Miguel
  • Ahedo García, Virginia
  • Arnaiz Rodríguez, Adrián
  • García Osorio, César
  • Santos Martín, José Ignacio
  • Galán Ordax, José Manuel
This paper applies complex network analysis to unveil the informal structure of the knowledge area of business organization — Organización de empresas— in Spain. To do so, we use the TESEO database. We retrieve and statically analyze all the theses referred to the UNESCO academic field of Organization and management of enterprises. Our results reveal a degree distribution of the participation in thesis examining committees and thesis supervision compatible with a truncated power law. Community analysis of the projected network of co-participation in thesis committees presents modular structure. When we focus on the backbone of such network, we find that the patterns detected can be partially explained by homophily of scholars that interact in the same academic association., Este artículo analiza la base de datos TESEO mediante redes complejas para obtener la estructura informal del área de conocimiento de Organización de Empresas en España. Para ello, toma todas las tesis referidas al campo UNESCO de Organización y Dirección de Empresas, y las analiza de forma estática. Los resultados revelan una distribución de grado de la participación en tribunales y supervisión de tesis compatible con una ley de potencia truncada, y una clara estructura modular. Los patrones encontrados analizando el núcleo de la red pueden ser parcialmente explicados por homofilia entre los investigadores de la misma asociación académica., The authors acknowledge financial support from the Spanish Ministry of Science, Innovation and Universities (HAR2017-90883-REDC and RED2018‐102518‐T), from the Junta de Castilla y León - Consejería de Educación (BDNS 425389 and BU055P20) and the predoctoral grant awarded to Virginia Ahedo (supported by the European Social Fund), and from NVIDIA Corporation and its donation of the TITAN Xp GPUs that facilitated this research. In addition, the authors would like to thank Dr. Manzanedo, Dra. SaizBárcena, Dr. Solé Parellada, Dr. Izquierdo and Dr. del Olmo for their insightful help to improve the manuscript and Alicia Olivares-Gil and Yi Peng Ji for some programming help.
Repositorio Institucional de la Universidad de Burgos (RIUBU)
Publicación Artículo científico (JournalArticle)

Semi-supervised classification with pairwise constraints: A case study on animal identification from video

  • Kuncheva, Ludmila I. .
  • Garrido Labrador, José Luis
  • Ramos Pérez, Ismael
  • Hennessey, Samuel L.
  • Rodríguez Diez, Juan José
Mainstream semi-supervised classification assumes that part of the available data are labelled. Here we assume that, in addition to the labels, we have pairwise constraints on the unlabelled data. Each constraint links two instances, and is one of Must Link (ML, belong to the same class) or Cannot Link (CL, belong to different classes). We propose an approach that uses the labelled data to train a classifier and then applies the ML and CL constraints in subsequent labelling. In our approach, a set of instances are labelled at the same time. Our case study is on animal re-identification. The dataset consists of five free-camera video clips of animals (koi fish, pigeons and pigs), annotated with bounding boxes and animal identities. The proposed approach combines the representations or classifiers predictions from the bounding boxes of consecutive frames. We demonstrate that our approach outperforms standard classifiers, constrained clustering, as well as inductive and transductive semi-supervised learning, using five feature representations., This work is supported by the UKRI Centre for Doctoral Training in Artificial Intelligence, Machine Learning and Advanced Computing (AIMLAC), funded by grant EP/S023992/1. This work is also supported by the Junta de Castilla León under project BU055P20 (JCyL/FEDER, UE), and the Ministry of Science and Innovation of Spain under the projects PID2020-119894GB-I00/AEI/10.13039/501100011033, co-financed through European Union FEDER funds. J.L. Garrido-Labrador is supported through Consejería de Educación of the Junta de Castilla y León and the European Social Fund through a pre-doctoral grant (EDU/875/2021). I. Ramos-Perez is supported by the predoctoral grant (BDNS 510149) awarded by the Universidad de Burgos, Spain. J.J. Rodríguez was supported by mobility grant PRX21/00638 of the Spanish Ministry of Universities .
Repositorio Institucional de la Universidad de Burgos (RIUBU)
Publicación Artículo científico (JournalArticle)

Ensemble methods and semi-supervised learning for information fusion: A review and future research directions

  • Garrido Labrador, José Luis
  • Serrano Mamolar, Ana
  • Maudes Raedo, Jesús M.
  • Rodríguez Diez, Juan José
  • García Osorio, César
Advances over the past decade at the intersection of information fusion methods and Semi-Supervised Learning (SSL) are investigated in this paper that grapple with challenges related to limited labelled data. To do so, a bibliographic review of papers published since 2013 is presented, in which ensemble methods are combined with new machine learning algorithms. A total of 128 new proposals using SSL algorithms for ensemble construction are identified and classified. All the methods are categorised by approach, ensemble type, and base classifier. Experimental protocols, pre-processing, dataset usage, unlabelled ratios, and statistical tests are also assessed, underlining the major trends, and some shortcomings of particular studies. It is evident from this literature review that foundational algorithms such as self-training and co-training are influencing current developments, and that innovative ensemble techniques are continuing to emerge. Additionally, valuable guidelines are identified in the review for improving research into intrinsically semi-supervised and unsupervised pre-processing methods, especially for regression tasks., This work was supported through the Junta de Castilla y León (JCyL) (regional goverment) under project BU055P20 (JCyL/FEDER, UE), the Spanish Ministry of Science and Innovation under project PID2020-119894GB-I00 co-financed through European Union FEDER funds, and project TED2021-129485B-C43 funded by MCIN/AEI/10.13039/501100011033 and the European Union NextGenerationEU/PRTR. J.L. Garrido-Labrador is supported through Consejería de Educación of the Junta de Castilla y León and the European Social Fund through a pre-doctoral grant EDU/875/2021 (Spain).
Repositorio Institucional de la Universidad de Burgos (RIUBU)
Publicación Contribución de congreso (ConferenceOutput)

Evaluation of Semi-Supervised Machine Learning applied to Affective State Detection

  • Martin-Melero, Íñigo
  • Serrano Mamolar, Ana
  • Rodríguez Diez, Juan José
Comunicación presentada en: EmotionAware 2024: Eighth International Workshop on Emotion Awareness for Pervasive Computing Beyond Traditional Approaches, held as part of the 2024 IEEE International Conference on Pervasive Computing and Communications Workshops and other Affiliated Events (PerCom Workshops), 11–15 March 2024, Biarritz, France., The affective computing field usually concerns data that is difficult, expensive or time-consuming to label. One way to overcome this limitation is the application of Semi-Supervised Machine Learning, that typically works with a small set of labeled data and a larger one of unlabeled data. This paper assesses the suitability of these techniques on the prediction of affective state, by analyzing the physiological and emotional response data of 30 different subjects while watching several emotion-eliciting videos. Three Semi-Supervised Learning algorithms are compared with their Supervised base classifiers in both a subject-independent and subject-dependent analyses, across a widely extended dataset. In view of the results, it can be concluded that Semi-Supervised Learning did not outperform their respective Supervised base classifiers for this specific dataset as it was expected. Subject-dependent classification resulted in accuracy rates between 68% and 85%, whereas the accuracy rates were between 38% and 72% for subject-independent classification., This work has been supported by the Junta de Castilla y Leon under project BU055P20 (JCyL/FEDER, UE), by the Ministry of Science and Innovation under project PID2020-119894GB-I00, co-financed through European Union FEDER funds, and by project TED2021-129485B-C43 funded by MCIN/AEI/10.13039/501100011033 and the European Union “NextGenerationEU”/PRTR.
Repositorio Institucional de la Universidad de Burgos (RIUBU)
Publicación Tesis doctoral (DoctoralThesis)

Machine Learning Approaches in Bioinformatics: Advances in Transcription and Protein Fitness Prediction

  • Barbero Aparicio, José Antonio
As we move deeper into the information age, bioinformatics has become increasingly important in modern biology, largely due to its critical role in processing and analyzing the vast amounts of complex data generated in the field. Traditional methods are often overwhelmed by the large volume and complexity of this data, positioning machine learning techniques as an optimal solution. Exploring the intersection of machine learning and bioinformatics offers numerous opportunities to develop and improve computational tools designed to handle and gain insights from these vast datasets. The main objective of this thesis is to develop a thorough exploration of the possibilities of machine learning in the field of bioinformatics, with a particular focus on specific problems such as transcription start and protein fitness prediction. Furthermore, given the similarities between bioinformatics sequence data and the natural language processing domain, the research emphasizes the use of sequence-based methods. Our research has resulted in several contributions to the field in the form of three scientific papers. The first two focus on transcription start prediction. In the first, we discovered that the integration of biophysical simulations in conjunction with the DNA sequence can improve the results of machine learning methods. Additionally, in our second paper we concluded that, while support vector machines have been a traditional choice for transcription start prediction, our research suggests that deep learning methods outperform them, marking a paradigm shift in the field. In addition, we presented custom-built datasets using Ensembl data, providing a valuable resource for future studies. The third paper addresses the issue of protein fitness prediction specifically in scarce dataset scenarios and concludes that deep transfer learning methods get established as the best alternative when compared with other strategies well suited for such situations, such as semi-supervised learning., A medida que nos seguimos adentrando en la era de la información, la bioinformática está pasando a ser cada vez más importante en la biología moderna, en gran parte debido a su papel crítico en el procesamiento y análisis de la gran cantidad de datos complejos generados en el campo. A menudo los métodos tradicionales encuentran dificultades debidas al gran volumen y complejidad de estos datos, posicionando a las técnicas de aprendizaje automático como una solución más óptima. La exploración de la intersección entre aprendizaje automático y la bioinformática ofrece numerosas oportunidades para el desarrollo y la mejora de herramientas computacionales diseñadas para manejar y obtener información crítica sobre estos grandes conjuntos de datos. El objetivo principal de esta tesis es el desarrollo de una exploración exhaustiva de las posibilidades del aprendizaje automático en bioinformática, con un enfoque específico en problemas como la predicción del inicio de la transcripción y la predicción del fitness en las proteínas. Además, dadas las similitudes entre las secuencias bioinformáticas y el campo del procesamiento del lenguaje natural, la investigación tiene un claro énfasis en el uso de métodos basados en secuencias. Este trabajo ha resultado en la producción de varias contribuciones al campo en forma de tres artículos científicos. Los dos primeros se centran en la predicción del inicio de la transcripción. En el primero de ellos descubrimos que la integración de simulaciones biofísicas en conjunto con la secuencia de ADN puede mejorar los resultados de los métodos de aprendizaje automático. En el segundo, además, llegamos a la conclusión de que, mientras que las máquinas de soporte vectorial han sido una opción muy establecida en el campo de la predicción del inicio de la transcripción, nuestra investigación sugiere que los métodos de aprendizaje profundo los superan, marcando un cambio de paradigma en el área. Además, presentamos conjuntos de datos personalizados a partir de datos de Ensembl, proporcionando un recurso valioso para futuros estudios. El tercer artículo aborda la predicción del fitness en proteínas, específicamente en escenarios con conjuntos de datos escasos y concluye que los métodos de deep transfer learning se establecen como la mejor alternativa ante otras estrategias bien adaptadas a tales situaciones, como los métodos de aprendizaje semi-supervisado., his thesis has been funded through a pre-doctoral grant by the University of Burgos. The work included in this thesis has also been supported by the Junta de Castilla y León under project BU055P20 (JCyL/FEDER, UE), by the Ministry of Science and Innovation under project PID2019-109481GB-I00 and the Junta de Andalucia under project UCO1264182, in both cases co-financed through European Union FEDER funds and by Fundación Bancaria la Caixa under project 2020/00062/001. NVIDIA Corporation donated the TITAN Xp GPUs used in this research.
Repositorio Institucional de la Universidad de Burgos (RIUBU)
Publicación Tesis doctoral (DoctoralThesis)

Técnicas de preprocesamiento en datos anchos, Preprocessing techniques for wide data

  • Ramos Pérez, Ismael
En el aprendizaje automático, especialmente en biología y genómica, los conjuntos de datos son llamados anchos, Los datos anchos contienen una gran cantidad de características pero pocas instancias. Esta alta dimensionalidad dificulta el aprendizaje de los modelos de aprendizaje, causando sobreajuste y sesgos. Para abordar estos problemas, las técnicas de preprocesamiento de datos son fundamentales. Esta tesis explora diversas estrategias de preprocesamiento, como la selección y reducción de características, y el remuestreo para equilibrar las clases. Además de técnicas semi-supervisadas. A través de tres artículos, se comparan diferentes estrategias de preprocesado y, en el campo de la selección de características en aprendizaje semi-supervisado y se propone una nueva taxonomía de algoritmos. Los resultados obtenidos demuestran la importancia de combinar adecuadamente las técnicas de preprocesamiento para obtener modelos más robustos y generales. Esta investigación proporciona una guía práctica para abordar los desafíos planteados por los datos anchos en el aprendizaje automático., The research carried out for the development of this doctoral thesis has been partially funded by the Junta de Castilla y León under project BU055P20 (JCyL/FEDER, UE), by the Spanish Ministry of Science and Innovation (projects PID2020-119894GB-I00 and TED 2021129485B-C43) and by “la Caixa“ Foundation, under agreement LCF/PR/PR18/51130007. The author has been recipient of the predoctoral grant by the Universidad de Burgos., Doctorado en Ingeniería y Tecnologías Industrial, Informática y Civil
Repositorio Institucional de la Universidad de Burgos (RIUBU)
Publicación Tesis doctoral (DoctoralThesis)

Optimization of industrial processes through semi-supervised learning

  • Ramírez Sanz, José Miguel
Enlas últimas décadas, el aprendizaje automático ha adquirido una gran popularidad, que ha ayudado a impulsar su ya creciente desarrollo. Dentro del aprendizaje automático, una nueva técnica llamada aprendizaje semisupervisado es capaz de aprovechar tanto los datos etiquetados comolos noetiquetados. Esto permite desarrollar modelos más precisos y con mejor capacidad de generalización gracias a la información que son capaces de extraer de los datos. La industria ha sido el motor del progreso humano, transformando recursos en bienes y productos de alta complejidad mediante la innovación. Mediante el desarrollo tecnológico y económico, la industria ha permitido que la sociedad avance hacia nuevos horizontes de productividad y calidad de vida. Esta tesis explora aplicaciones del aprendizaje semisupervisado en problemas industriales, con el objetivo de comprender su estado del arte, investigar nuevas metodologías, diseñar modelos y aplicarlos a casos reales en la industria. Las aportaciones más destacables de esta tesis doctoral han sido revisar el estado del arte de la aplicación de métodos de aprendizaje semisupervisado aladetección ydiagnóstico defallos industriales. Se ha investigado y propuesto una nueva taxonomía para una nueva técnica desarrollada a partir del aprendizaje semisupervisado: el aprendizaje semisupervisado seguro. Además, se ha realizado el desarrollo de una versión semisupervisada del método más conocido del grupo de investigación, el Rotation Forest. También se ha elaborado el desarrollo y la aplicación de un método interpretable de selección de características basado en árboles para el problema de selección del método de transporte. Por último, se han aplicado y comparado modelos supervisados y semisupervisados al problema de selección del método de transporte. En resumen, esta tesis doctoral proporciona una contribución significativa al campo del aprendizaje semisupervisado aplicado a la industria, proponiendo nuevas metodologías y modelos que optimizan problemas industriales, validandolos en casos de estudio reales., In recent decades, machine learning has gained great popularity, which has helped boost its already growing development. Within machine learning, a new technique called semi-supervised learning is capable of leveraging both labeled and unlabeled data. This allows it to develop more precise models with better generalization capabilities due to the information they can extract from the data. Industry has been the engine of human progress, transforming resources into complex goods and products through innovation. Technological and economic development has enabled society to advance toward new horizons of productivity and quality of life. This doctoral thesis explores the applications of semi-supervised learning to industrial problems. The main objectives include understanding the state of the art of the applications of this technique to industrial problems, investigating new methodologies developed from this approach, designing new models of this type of learning, and applying them to real-world industrial environments. The main contributions of this doctoral thesis include reviewing the state of the art of semi-supervised learning methods applied to fault detection and diagnosis in industrial settings, highlighting the formulation of a series of best practices to improve future research in this field. Additionally, a new taxonomy has been investigated and proposed for a new technique derived from semi-supervised learning: safe semi-supervised learning. Also the development of a semi-supervised version of the most well-known method from the research group, the Rotation Forest, was also achieved. Moreover, a tree-based interpretable feature selection method was developed and applied to the problem of travel mode choice. Finally, supervised and semi-supervised models were applied and compared for the travel mode choice problem. In summary, this doctoral thesis makes a significant contribution to the field of semi-supervised learning applied to industry, proposing new methodologies and models that optimize industrial problems, validated through real-world case studies., This thesis was supported by the Junta de Castilla y León under project BUR055P20 (JCyL/FEDER, UE), the Ministry of Science and Innovation of Spain under the projects PID2020-119894GB-I00, TED2021129485B-C43 and PID2023-150694OA-I00, co-financed through European Union FEDERfunds. It also was supported through the Consejería de Educación of the Junta de Castilla y León and the European Social Fund through a pre-doctoral grant (EDU/875/2021)., Doctorado en Ingeniería y Tecnologías Industrial, Informática y Civil