Predicci贸n del riesgo de dimisi贸n de un empleado mediante t茅cnicas de Machine Learning y an谩lisis de redes

SEMINARIO DEL DOCTORADO DE ESTAD脥STICA, OPTIMIZACI脫N Y MATEM脕TICA APLICADA. CURSO 2016-2017

T铆tulo:聽Predicci贸n del riesgo de dimisi贸n de un empleado mediante t茅cnicas de Machine Learning y an谩lisis de redes

Ponentes:聽Guillem Duran (Colaborador con United Arab Emirates University (UAEU) y el grupo de investigaci贸n CIO 鈥淪istemas din谩micos y aplicaciones鈥)

Fecha: Jueves 20 de julio de 2017 a las聽12:30 horas

Lugar: Sala de seminarios, Edificio Torretamarit, Universidad Miguel Hern谩ndez聽(Campus de Elche)

Resumen:

La obtenci贸n de m茅tricas para predecir si un empleado va a dimitir de su actual puesto de trabajo es un tema de inter茅s para los departamentos de recursos humanos, debido al coste econ贸mico que conlleva una baja inesperada en la plantilla de una empresa. En esta charla analizamos c贸mo construir modelos predictivos del riesgo de dimisi贸n mediante t茅cnicas de Machine Learning y an谩lisis de redes. Utilizando los datos de la aplicaci贸n de MyHappyforce, dise帽ada para obtener comentarios de los empleados de una empresa, explicaremos el proceso a seguir para construir modelos predictivos del riesgo de dimisi贸n a partir de las interacciones entre los empleados. Durante esta charla empezaremos explicando en qu茅 consiste nuestro dataset, c贸mo debe de ser procesado, y cu谩les son las caracter铆sticas m谩s relevantes que pueden observarse al analizar los datos. A continuaci贸n, explicaremos c贸mo se pueden utilizar t茅cnicas de an谩lisis de redes para crear rasgos que permitan cuantificar num茅ricamente las interacciones entre los empleados, haciendo especial hincapi茅 en la t茅cnica de extracci贸n de grupos llamada Non-negative Matrix Factorization (NMF). Finalmente, analizaremos qu茅 caracter铆sticas son las que m谩s influyen a la hora de predecir el riesgo que un empleado tiene de abandonar su puesto de trabajo.

Categories: Sin categor铆a Tags:
Viernes, 14 de julio de 2017 Sin comentarios

Achieving a sustainable cost efficient business model in banking: The case of European banks

SEMINARIO DEL DOCTORADO DE ESTAD脥STICA, OPTIMIZACI脫N Y MATEM脕TICA APLICADA. CURSO 2016-2017

T铆tulo:聽Achieving a sustainable cost efficient business model in banking: The case of European banks

Ponentes: Ana Lozano鈥怴ivas (University of Malaga, Spain)

Fecha: Viernes聽14 de julio de 2017 a las聽10:00 horas

Lugar: Sala de seminarios, Edificio Torretamarit, Universidad Miguel Hern谩ndez聽(Campus de Elche)

Resumen:

Although the business model (BM) is the most fundamental task of the bank鈥檚 management to ensure sustained operation and profitability of a bank, it has become a subject of supervisor鈥檚 scrutiny due to the recent financial crisis when failing banks were rescued with public funds and supervisors were criticized around the world. Since one of the reasonsfor the financial crisis wasthatsome banks had (and still have) unsustainable BMs, sustainable BMs are, for example, on the top of the ECB鈥檚 agenda. Given its relevance, it is important to understand implication that BM characteristics have for bank performance in general and for cost efficiency in particular. Optimizing operating efficiency has become a necessity for the survival of bank. This is one of the top priorities for a bank, especially during times when revenue鈥恎enerating opportunities are sub鈥恛ptimal. This paper usesthe Herfindahl index to measure how concentrate the bank isin items of the asset, funding or income portfolio. We analyze efficiency of bank BM along three business dimensions, viz., assets, funding and income, for the European Banking Industry. We apply recently developed four component heteroskedastic cost model to investigate effects of three business dimensions to time鈥 varying bank cost inefficiency while controlling for bank effects and persistent cost inefficiency. In the proposed model we assume bank鈥恠pecific effects and persistent cost inefficiency random and distributed independently and identically across banks but time鈥恦arying cost inefficiency and noise terms are made heteroscedastic in terms of assets, funding and income diversification for each bank. Note that we are interpreting heteroscedasticity of the noise term as risk thereby meaning whether different forms of diversifications are risk enhancing or risk reducing.

Categories: Sin categor铆a Tags:
Mi茅rcoles, 12 de julio de 2017 Sin comentarios

Algunos valores para juegos TU como funciones de particiones

SEMINARIO DEL DOCTORADO DE ESTAD脥STICA, OPTIMIZACI脫N Y MATEM脕TICA APLICADA. CURSO 2016-2017

T铆tulo:聽Algunos valores para juegos TU como funciones de particiones

Ponentes: Gloria Fiestras-Janeiro (Universidad de Vigo)

Fecha: Martes聽18 de julio de 2017 a las聽12:30 horas

Lugar: Sala de seminarios, Edificio Torretamarit, Universidad Miguel Hern谩ndez聽(Campus de Elche)

Resumen:

Los juegos TU como funciones de particiones son estructuras apropiadas para modelizar las ganancias/costes de una coalici贸n teniendo en cuenta como se organizan el resto de los jugadores. En esta charla revisaremos algunos valores propuestos en la literatura, centr谩ndonos en sus propiedades. En particular, revisaremos el valor de Myerson (Myerson 1977), el valor de Bolger (Bolger 1989), el valor libre de externalidades (de Clippel y Serrano 2008; Do and Norde 2007) y el valor de McQuillin (2009)

Categories: Sin categor铆a Tags:
Mi茅rcoles, 28 de junio de 2017 Sin comentarios

Measuring Productivity Change Accounting for Adjustment Costs: Evidence from the Food Industry in the European Union

SEMINARIO DEL DOCTORADO DE ESTAD脥STICA, OPTIMIZACI脫N Y MATEM脕TICA APLICADA. CURSO 2016-2017

T铆tulo:聽Measuring Productivity Change Accounting for Adjustment Costs: Evidence from the Food Industry in the European Union

Ponentes: Magdalena Kapelko (Wroclaw University of Economics, Institute of Applied Mathematics, Department of Logistics)

Fecha: Martes 11 de julio de 2017 a las聽10:00 horas

Lugar: Sala de seminarios, Edificio Torretamarit, Universidad Miguel Hern谩ndez聽(Campus de Elche)

Resumen:

This paper extends the measurement of dynamic productivity change over time to provide its full decomposition into economically meaningful components in the Data Envelopment Analysis framework. The dynamic approach accounts for dynamics of production decisions via adjustment costs and is visualized as a dynamic Luenberger productivity change indicator. The paper also estimates the dynamic productivity change and its components for a large dataset of European food companies from 2004 till 2012, grouped into Eastern, Southern, and Western regions. The study reveals three main results. First, the overall trend of dynamic technical regress and positive dynamic technical inefficiency change across almost all regions and sectors was found. Second, some differences for this general pattern were found for the bakery industry and for Eastern European firms. Thirdly, there are also some remarkable changes in indicators observed during the periods related to the financial crisis and the volatility of agricultural commodity prices.

Categories: Sin categor铆a Tags:
Mi茅rcoles, 28 de junio de 2017 Sin comentarios

Aplicaciones de modelos jer谩rquicos bayesianos en Epidemiolog铆a en Plantas

SEMINARIO DEL DOCTORADO DE ESTAD脥STICA, OPTIMIZACI脫N Y MATEM脕TICA APLICADA. CURSO 2016-2017

T铆tulo:聽Aplicaciones de modelos jer谩rquicos bayesianos en Epidemiolog铆a en Plantas

Ponentes:聽Joaqu铆n Mart铆nez Minaya (Universitat de Val猫ncia)

Fecha: Viernes 16 de junio de 2017 a las聽11:30 horas

Lugar: Sala de seminarios, Edificio Torretamarit, Universidad Miguel Hern谩ndez聽(Campus de Elche)

Resumen:

En los 煤ltimos a帽os, el uso de modelos estad铆sticos complejos en disciplinas como la Epidemiolog铆a en plantas ha aumentado con el objetivo de ayudarnos a entender el comportamiento de las enfermedades en este 谩mbito. La dificultad que presentan algunos de estos modelos hace dif铆cil tanto el proceso de estimaci贸n como el de predicci贸n. Los modelos jer谩rquicos bayesianos parecen una buena alternativa para abordar este tipo de problemas y est谩n basados principalmente en la premisa de que ambos, informaci贸n e incertidumbre se pueden expresar en t茅rminos de distribuciones de probabilidad. En esta charla, presentamos modelizaciones para el estudio de algunas enfermedades en plantas, como son la mancha negra de los c铆tricos o la mancha foliar del caqui.

Categories: Sin categor铆a Tags:
Mi茅rcoles, 14 de junio de 2017 Sin comentarios

Small-scale fisheries in Brazil: a not-so-invisible problem

SEMINARIO DEL DOCTORADO DE ESTAD脥STICA, OPTIMIZACI脫N Y MATEM脕TICA APLICADA. CURSO 2016-2017

T铆tulo:聽Small-scale fisheries in Brazil: a not-so-invisible problem

Ponentes:聽Priscila Lopes (Universidade Federal do Rio Grande do Norte, Brazil)

Fecha: Viernes 16 de junio de 2017 a las 12:00聽horas

Lugar: Sala de seminarios, Edificio Torretamarit, Universidad Miguel Hern谩ndez聽(Campus de Elche)

Resumen:

Brazil is proportionally a relatively unimportant player in the world fisheries scenario. However, small-scale fisheries (SSF) specifically are crucial for local livelihoods and the food security of millions of people, both on the coast and inland. Such fisheries provide more than 50% of the total Brazilian catches, suggesting that they can be highly impacting on the local ecosystems, especially on vulnerable ones, such as coastal reefs. Despite having some preferential target species, SSF is highly diverse and adaptive, depending on the environment, but also on the socioeconomic context. For being mostly coastal and exploiting shallow waters, many of the stocks targeted cannot undergo intensive fishing pressure, which have put many key species in a vulnerable situation, with some of them being locally extinct already. Topping up the problem, Brazil does not have a fishery statistical program since 2011, and has been cutting research funding continuously since 2014. To partially overcome the limitations, we have been developing alternative, affordable and reliable tools to collect and reconstruct fishing data, usually with the support of fishers. As there is official interest in managing such fisheries either, we have to stimulate and count on participatory initiatives to implement management as well and to understand, counterbalance or adjust, when necessary, the general impacts of development projects and of protected areas.

Categories: Sin categor铆a Tags:
Jueves, 8 de junio de 2017 Sin comentarios

A convex approach to differential inclusions with prox-regular sets

SEMINARIO DEL DOCTORADO DE ESTAD脥STICA, OPTIMIZACI脫N Y MATEM脕TICA APLICADA. CURSO 2016-2017

T铆tulo:聽A convex approach to differential inclusions with prox-regular sets

Ponentes: Dr. Abderrahim Hantoute (Center for Mathematical Modeling, University of Chile)

Fecha: Martes 13 de junio de 2017 a las聽18:30 horas

Lugar: Sala de seminarios, Edificio Torretamarit, Universidad Miguel Hern谩ndez聽(Campus de Elche)

Resumen:

We study the existence and stability of the solutions for differential inclusions governed by the normal cone to a prox-regular set and subject to a Lipschitz perturbation. We prove that such apparently more general systems can be indeed remodeled into the classical theory of differential inclusions involving maximal monotone operators. This result permits us to make use of the rich and abundant achievements in this class of monotone operators to derive the desired existence result and stability analysis, as well as the continuity and differentiability properties of the solutions. This going back and forth between these two kinds of differential inclusions is made possible thanks to a viability result for maximal monotone operators. This is a joint work with S. Adly and Bao Tran Nguyen.

 

Categories: Sin categor铆a Tags:
Jueves, 8 de junio de 2017 Sin comentarios

Differences in teachers鈥 efficiency: evidence for Spain

SEMINARIO DEL DOCTORADO DE ESTAD脥STICA, OPTIMIZACI脫N Y MATEM脕TICA APLICADA. CURSO 2016-2017

T铆tulo:聽Differences in teachers鈥 efficiency: evidence for Spain

Ponentes: Gabriela Sicilia (Universidad Aut贸noma de Madrid)

Fecha: Mi茅rcoles聽7 de junio de 2017 a las聽12:00 horas

Lugar: Sala de seminarios, Edificio Torretamarit, Universidad Miguel Hern谩ndez聽(Campus de Elche)

Resumen:

This research contributes to the ongoing debate about differences in teachers鈥 performance. We introduce a new methodology that combine production frontier and impact evaluation insights that allows using DEA as an identification strategy of a treatment with high and low quality teachers within schools to assess their performance. We use a unique database in Spain that supplies information on two classrooms at 4th grade in primary education, where students and teachers were randomly assigned into the two classrooms at each school. We find considerable differences in teachers鈥 efficiency across schools and also significant effects on students鈥 achievement. In line with previous findings, we find that neither teacher experience nor academic training explains teachers鈥 efficiency. Conversely, being a female teacher, having worked five or more years in the same school, repeating a second year with the same group or having smaller class sizes positively affects the efficiency of teachers.

Categories: Sin categor铆a Tags:
Lunes, 5 de junio de 2017 Sin comentarios

A Wald-type test statistic based on robust modified median estimator in logistic regression models

SEMINARIO DEL DOCTORADO DE ESTAD脥STICA, OPTIMIZACI脫N Y MATEM脕TICA APLICADA. CURSO 2016-2017

T铆tulo:聽A Wald-type test statistic based on robust modified median estimator in logistic regression models

Ponentes: Tom谩拧 Hobza (Czech Technical University in Prague)

Fecha: Lunes 5 de junio de 2017 a las聽12:00 horas

Lugar: Sala de seminarios, Edificio Torretamarit, Universidad Miguel Hern谩ndez聽(Campus de Elche)

Resumen:

This contribution deals with robust estimation and hypothesis testing in logistic regression models. It introduces a modified median estimator of the underlying parameters of these models based on statistically smoothed binary responses and gives its asymptotic distribution.

Using the modified median estimator, a Wald-type test statistic for testing linear hypotheses in the logistic regression model is also considered and its asymptotic distribution is obtained.

Sensitivity to contaminations and leverage points is studied by simulations and compared in this manner with the sensitivity of some robust estimators and tests previously introduced to the logistic regression. The new estimator and test based on it appears to be more robust for larger sample sizes and higher levels of contamination.

Key words:聽Logistic regression, Median estimator, Robustness, Wald-type tets.

Categories: Sin categor铆a Tags:
Lunes, 5 de junio de 2017 Sin comentarios

ANUNCIO: INTRODUCCI脫N AL DESARROLLO DE PAQUETES EN R

CURSO

T铆tulo: Introducci贸n al desarrollo de paquetes en R

Ponentes:

Virgilio G贸mez-Rubio (Universidad Castilla-La Mancha)

Fecha:聽11-12 julio 2017

Lugar:聽Facultad de CC de la Empresa.聽Universidad Polit茅cnica de Cartagena

Informaci贸n:

Cartel Taller R

Categories: Sin categor铆a Tags:
Lunes, 5 de junio de 2017 Sin comentarios