Person: Portela García-Miguel, Javier
Loading...
First Name
Javier
Last Name
Portela García-Miguel
Affiliation
Universidad Complutense de Madrid
Faculty / Institute
Estudios estadísticos
Department
Estadística y Ciencia de los Datos
Area
Estadística e Investigación Operativa
Identifiers
19 results
Search Results
Now showing 1 - 10 of 19
Item Aplicación de métodos numéricos en inferencia bayesiana : implementación de un método bayesiano robusto(2004) Portela García-Miguel, Javier; Gómez-Villegas, Miguel ÁngelEl objetivo de este trabajo es desarrollar técnicas para la aplicación de la familia de distribuciones Potencial Exponencial, dentro del marco de la Inferencia Bayesiana, con especial incidencia en el problema de selección de modelos bayesianos. En particular, se presenta una generalización de esta familia, y se desarrollan un método Monte-Carlo, un método de simulación vía muestreo de Gibbs y un método de simulación que utiliza una representación en mixturas de esta familia, para establecer inferencias sobre las distribuciones a posteriori surgidas del planteamiento bayesiano. A través del parámetro de control de curtosis puede plantearse un contraste bayesiano de hipótesis nula puntual para contrastar normalidad de los datos en el marco de esta familia. Se plantea este contraste desde un enfoque basado en medidas de discrepancia, presentando una medida basada en el cálculo de regiones de máxima densidad a posteriori y haciendo un estudio de simulación. Finalmente se aplican las técnicas desarrolladas anteriormente en el marco de modelos bayesianos, concretamente en modelos lineales, modelos no lineales, y modelos longitudinales, poniendo de relieve el interés de la utilización de esta familia en problemas de robustez en modelos bayesianosItem Estimation of Anonymous Email Network Characteristics through Statistical Disclosure Attacks(Sensors, 2016) Portela García-Miguel, Javier; García Villalba, Luis Javier; Silva Trujillo, Alejandra Guadalupe; Sandoval Orozco, Ana Lucila; Kim, Tai-HoonSocial network analysis aims to obtain relational data from social systems to identify leaders, roles, and communities in order to model profiles or predict a specific behavior in users’ network. Preserving anonymity in social networks is a subject of major concern. Anonymity can be compromised by disclosing senders’ or receivers’ identity, message content, or sender-receiver relationships. Under strongly incomplete information, a statistical disclosure attack is used to estimate the network and node characteristics such as centrality and clustering measures, degree distribution, and small-world-ness. A database of email networks in 29 university faculties is used to study the method. A research on the small-world-ness and Power law characteristics of these email networks is also developed, helping to understand the behavior of small email networks.Item Multiple scenarios of quality of life index using fuzzy linguistic quantifiers: the case of 85 countries in Numbeo(Mathematics, 2022) Shu, Ziwei; Carrasco González, Ramón Alberto; Portela García-Miguel, Javier; Manuel Sánchez-Montañés, Manuel; Chiclana, FranciscoIn economic development, in addition to comparing the gross domestic product (GDP) between nations, it is critical to assess the quality of life to gain a holistic perspective of their different aspects. However, the quality of life index (QOLI) is a subjective term that can be difficult to quantify. Although this composite index is typically calculated using universal weights proposed by experts to aggregate indicators, such as safety indexes, healthcare indexes, pollution indexes, and housing indicators, it is complicated to balance multiple dimensions whose weights are adjusted to account for different countries’ circumstances. Therefore, this paper aims to construct various scenarios of the QOLI, using linguistic quantifiers of the ordered weighted averaging (OWA) operator, and the 2-tuple linguistic model. Numbeo, one of the largest quality of life information databases, was used in this paper to estimate the QOLI in 85 countries. Uncertainty and sensitivity analyses were employed to assess the robustness of the QOLI. The results of the proposed model are compared with those obtained using the Numbeo formulation. The results show that the proposed model increases the linguistic interpretability of the QOLI, and obtains different QOLIs, based on diverse country contexts.Item Estimación de la tasa de retorno de la carta del censo de los Estados Unidos a través del modelo de regresión lineal y técnicas de predicción inteligentes.(2014) Jiménez-Moro, José Luís; Portela García-Miguel, JavierEn este proyecto se ha realizado un estudio de la tasa de respuesta que obtiene el censo de los Estados Unidos con la carta que envía a sus ciudadanos. En el estudio se han utilizado tres técnicas de predicción distintas valorando el error cuadrático medio en cada una de ellas. Los resultados más precisos se han obtenido con una Red Neuronal, seguido del algoritmo Gradient Boosting y el modelo de regresión lineal. La forma de valorar la precisión de las técnicas predictivas se ha basado no solo en el cálculo del error cuadrático medio sino que además, éste ha sido representado sobre un mapa del país permitiendo ver la mejora en precisión en función de la técnica utilizada.Item Learning strategies for sensitive content detection(Electronics, 2023) Povedano Álvarez, Daniel; Sandoval Orozco, Ana Lucila; Portela García-Miguel, Javier; García Villalba, Luis Javier; Guo, ZhenhuaCurrently, the volume of sensitive content on the Internet, such as pornography and child pornography, and the amount of time that people spend online (especially children) have led to an increase in the distribution of such content (e.g., images of children being sexually abused, real-time videos of such abuse, grooming activities, etc.). It is therefore essential to have effective IT tools that automate the detection and blocking of this type of material, as manual filtering of huge volumes of data is practically impossible. The goal of this study is to carry out a comprehensive review of different learning strategies for the detection of sensitive content available in the literature, from the most conventional techniques to the most cutting-edge deep learning algorithms, highlighting the strengths and weaknesses of each, as well as the datasets used. The performance and scalability of the different strategies proposed in this work depend on the heterogeneity of the dataset, the feature extraction techniques (hashes, visual, audio, etc.) and the learning algorithms. Finally, new lines of research in sensitive-content detection are presented.- Project number: 156
Item Análisis de los contenidos docentes de matemáticas en el doble grado ADE-Informática(2018) García Pineda, M. Pilar; Heras Martínez, Antonio José; Blanco García, Susana; Balbas Aparicio, Raquel; García Villalba, Luis Javier; Sandoval Orozco, Ana Lucila; Portela García-Miguel, Javier; Riomoros Callejo, María Isabel; Rebollo Castillo, Francisco Javier Item Estudio de la percepción pública de la vacuna contra la COVID-19 mediante técnicas de PLN y de aprendizaje automático(2021) Povedano Álvarez, Daniel; Portela García-Miguel, Javier; Armas Vega, Esteban AlejandroLa pandemia de COVID-19 causada por el nuevo coronavirus SARS-CoV-2 ha tenido un impacto significativo en la sociedad, tanto por los graves efectos sanitarios y económicos como por los efectos de las medidas sanitarias para evitar su propagación. Gracias a las técnicas de PLN se ha podido investigar las actitudes hacia la vacunación, siendo particularmente oportuno en estos momentos ante la llegada de las vacunas contra la COVID-19. Este trabajo tiene una doble finalidad, por un lado estudiar la percepción hacia la vacunación contra la COVID-19, mediante técnicas de Procesamiento de Lenguaje Natural y por otro, la construcción de un clasificador de sentimientos interpretable mediante técnicas de Aprendizaje Automático. Para ello se uitlizaron 4.000.000 de tweets relacionados con la vacunación en el periodo comprendido entre el 15 de noviembre de 2020 y el 16 de diciembre de 2020 como conjunto de datos. El periodo de tiempo seleccionado es esencial porque durante este tiempo se publicaron los primeros resultados de las vacunas contra la COVID-19, como Pfizer y Moderna, surgieron un debate público. El análisis de la percepción sugiere que hay un número significativo de tweets negativos que pueden poner en peligro el objetivo de alcanzar la inmunidad de rebaño. En cuanto a los resultados del clasificador de sentimientos multiclase se obtuvo un 92% de ROC-AUC con el algoritmo LinearSVC.Item A model integrating the 2‑tuple linguistic model and the CRITIC‑AHP method for hotel classification(SN Computer Science, 2023) Shu, Ziwei; Carrasco González, Ramón Alberto; Portela García-Miguel, Javier; Sánchez‑Montañés, Manuel; Pal, Umapada; Yuen, ChauHotel classification is essential for hotel managers and customers. It can assist hotel managers in better understanding the needs of their customers and in improving various aspects of the hotel through relevant strategies. It also aids customers in choosing appropriate accommodations according to their preferences regarding hotel location, services, and other aspects. This paper aims to improve our previous model by incorporating expert opinions into the weight calculation, thereby increasing its practical applicability. The extended model combines the analytical hierarchy process (AHP) and the CRiteria Importance Through Intercriteria Correlation (CRITIC) methods, introducing a novel approach for calculating the weights of each aspect. The 2-tuple linguistic model is retained in the extended model to resolve the problem of information loss in linguistic information fusion. Finally, various hotel segments are obtained with the weighted K-means clustering. A dataset with over fifty million hotel reviews from TripAdvisor has been applied to evaluate the extended model. The results show that the extended model achieves denser and better separated hotel clusters than our previous model, while maintaining the same advantages. This model is more likely to help hotel managers create better strategies to tackle hotel weaknesses or gain competitive advantages, as it combines two types of weights that improve clustering results: the quantity of information in each hotel aspect and the expert judgment of each aspect's importance in hotel development.Item Differentiated models in the collaborative transport economy: a mixture analysis for Blablacar and Uber(2021) Quirós Romero, Cipriano; Portela García-Miguel, Javier; Marín Sanz, Raquel; Dabić, Marina; Griffy-Brown, CharlaThe collaborative economy has become one of the fastest-growing areas since the great recession of 2008, with passenger transport activities being of particular importance. Based on on-demand ride services, in which passengers are connected with community drivers through a smartphone app, these collaborative transport activities are not all of a homogeneous nature, and need to be differentiated according to the type of service provided to understand the socioeconomic and environmental characteristics that determine their widespread use. The objective of this paper is to analyze the determinants of use of the two main online platforms for transport in Europe: Blablacar (ride-sharing) and Uber (ride-hailing), representing two different business models for shared mobility. Our contribution is an empirical analysis emphasizing the role of digital skills and employing a mixture analysis technique that permits the distinction between these two platforms models. Our results show that beyond their strong linkage to the digital environment, the determinants of using both platforms differ. In the case of Blablacar, the results show a closed user profile influenced by age, educational level, live as a couple or urban environment. However, the user profile for Uber is open and more diffuse, with a preference for female population.Item