On the impact of disproportional samples in credit scoring models: An application to a Brazilian bank data

Statistical methods have been widely employed to assess the capabilities of credit scoring classification models in order to reduce the risk of wrong decisions when granting credit facilities to clients. The predictive quality of a classification model can be evaluated based on measures such as sensitivity, specificity, predictive values, accuracy, correlation coefficients and information theoretical measures, such as relative entropy and mutual information. In this paper we analyze the performance of a naive logistic regression model (Hosmer & Lemeshow, 1989) and a logistic regression with state-dependent sample selection model (Cramer, 2004) applied to simulated data. Also, as a case study, the methodology is illustrated on a data set extracted from a Brazilian bank portfolio. Our simulation results so far revealed that there is no statistically significant difference in terms of predictive capacity between the naive logistic regression models and the logistic regression with state-dependent sample selection models. However, there is strong difference between the distributions of the estimated default probabilities from these two statistical modeling techniques, with the naive logistic regression models always underestimating such probabilities, particularly in the presence of balanced samples. (C) 2012 Elsevier Ltd. All rights reserved.

Structure learning of graphical models for task-oriented robot grasping

In the collective imaginaries a robot is a human like machine as any androids in science fiction. However the type of robots that you will encounter most frequently are machinery that do work that is too dangerous, boring or onerous. Most of the robots in the world are of this type. They can be found in auto, medical, manufacturing and space industries. Therefore a robot is a system that contains sensors, control systems, manipulators, power supplies and software all working together to perform a task. The development and use of such a system is an active area of research and one of the main problems is the development of interaction skills with the surrounding environment, which include the ability to grasp objects. To perform this task the robot needs to sense the environment and acquire the object informations, physical attributes that may influence a grasp. Humans can solve this grasping problem easily due to their past experiences, that is why many researchers are approaching it from a machine learning perspective finding grasp of an object using information of already known objects. But humans can select the best grasp amongst a vast repertoire not only considering the physical attributes of the object to grasp but even to obtain a certain effect. This is why in our case the study in the area of robot manipulation is focused on grasping and integrating symbolic tasks with data gained through sensors. The learning model is based on Bayesian Network to encode the statistical dependencies between the data collected by the sensors and the symbolic task. This data representation has several advantages. It allows to take into account the uncertainty of the real world, allowing to deal with sensor noise, encodes notion of causality and provides an unified network for learning. Since the network is actually implemented and based on the human expert knowledge, it is very interesting to implement an automated method to learn the structure as in the future more tasks and object features can be introduced and a complex network design based only on human expert knowledge can become unreliable. Since structure learning algorithms presents some weaknesses, the goal of this thesis is to analyze real data used in the network modeled by the human expert, implement a feasible structure learning approach and compare the results with the network designed by the expert in order to possibly enhance it.

On the Behaviour of Marginal and Conditional Akaike Information Criteria in Linear Mixed Models

In linear mixed models, model selection frequently includes the selection of random effects. Two versions of the Akaike information criterion (AIC) have been used, based either on the marginal or on the conditional distribution. We show that the marginal AIC is no longer an asymptotically unbiased estimator of the Akaike information, and in fact favours smaller models without random effects. For the conditional AIC, we show that ignoring estimation uncertainty in the random effects covariance matrix, as is common practice, induces a bias that leads to the selection of any random effect not predicted to be exactly zero. We derive an analytic representation of a corrected version of the conditional AIC, which avoids the high computational cost and imprecision of available numerical approximations. An implementation in an R package is provided. All theoretical results are illustrated in simulation studies, and their impact in practice is investigated in an analysis of childhood malnutrition in Zambia.

Analysis of postprandial lipemia as a Cardiovascular Disease risk factor using genetic and clinical information: an Artificial Neural Network perspective

Clinical studies indicate that exaggerated postprandial lipemia is linked to the progression of atherosclerosis, leading cause of Cardiovascular Diseases (CVD). CVD is a multi-factorial disease with complex etiology and according to the literature postprandial Triglycerides (TG) can be used as an independent CVD risk factor. Aim of the current study is to construct an Artificial Neural Network (ANN) based system for the identification of the most important gene-gene and/or gene-environmental interactions that contribute to a fast or slow postprandial metabolism of TG in blood and consequently to investigate the causality of postprandial TG response. The design and development of the system is based on a dataset of 213 subjects who underwent a two meals fatty prandial protocol. For each of the subjects a total of 30 input variables corresponding to genetic variations, sex, age and fasting levels of clinical measurements were known. Those variables provide input to the system, which is based on the combined use of Parameter Decreasing Method (PDM) and an ANN. The system was able to identify the ten (10) most informative variables and achieve a mean accuracy equal to 85.21%.

Neural network based glucose - insulin metabolism models for children with Type 1 diabetes

In this paper two models for the simulation of glucose-insulin metabolism of children with Type 1 diabetes are presented. The models are based on the combined use of Compartmental Models (CMs) and artificial Neural Networks (NNs). Data from children with Type 1 diabetes, stored in a database, have been used as input to the models. The data are taken from four children with Type 1 diabetes and contain information about glucose levels taken from continuous glucose monitoring system, insulin intake and food intake, along with corresponding time. The influences of taken insulin on plasma insulin concentration, as well as the effect of food intake on glucose input into the blood from the gut, are estimated from the CMs. The outputs of CMs, along with previous glucose measurements, are fed to a NN, which provides short-term prediction of glucose values. For comparative reasons two different NN architectures have been tested: a Feed-Forward NN (FFNN) trained with the back-propagation algorithm with adaptive learning rate and momentum, and a Recurrent NN (RNN), trained with the Real Time Recurrent Learning (RTRL) algorithm. The results indicate that the best prediction performance can be achieved by the use of RNN.

EFFECTS OF INFORMATION DISPLAY ON THE CONSTRUCTION OF CLINICIAN MENTAL MODELS

Objective: To determine how a clinician’s background knowledge, their tasks, and displays of information interact to affect the clinician’s mental model. Design: Repeated Measure Nested Experimental Design Population, Sample, Setting: Populations were gastrointestinal/internal medicine physicians and nurses within the greater Houston area. A purposeful sample of 24 physicians and 24 nurses were studied in 2003. Methods: Subjects were randomized to two different displays of two different mock medical records; one that contained highlighted patient information and one that contained non-highlighted patient information. They were asked to read and summarize their understanding of the patients aloud. Propositional analysis was used to understand their comprehension of the patients. Findings: Different mental models were found between physicians and nurses given the same display of information. The information they shared was very minor compared to the variance in their mental models. There was additionally more variance within the nursing mental models than the physician mental models given different displays of the same information. Statistically, there was no interaction effect between the display of information and clinician type. Only clinician type could account for the differences in the clinician comprehension and thus their mental models of the cases. Conclusion: The factors that may explain the variance within and between the clinician models are clinician type, and only in the nursing group, the use of highlighting.

Aplicación de técnicas de inteligencia artificial a la predicción de contaminantes atmosféricos

Abstract Air pollution is a big threat and a phenomenon that has a specific impact on human health, in addition, changes that occur in the chemical composition of the atmosphere can change the weather and cause acid rain or ozone destruction. Those are phenomena of global importance. The World Health Organization (WHO) considerates air pollution as one of the most important global priorities. Salamanca, Gto., Mexico has been ranked as one of the most polluted cities in this country. The industry of the area led to a major economic development and rapid population growth in the second half of the twentieth century. The impact in the air quality is important and significant efforts have been made to measure the concentrations of pollutants. The main pollution sources are locally based plants in the chemical and power generation sectors. The registered concerning pollutants are Sulphur Dioxide (SO2) and particles on the order of ∼10 micrometers or less (PM10). The prediction in the concentration of those pollutants can be a powerful tool in order to take preventive measures such as the reduction of emissions and alerting the affected population. In this PhD thesis we propose a model to predict concentrations of pollutants SO2 and PM10 for each monitoring booth in the Atmospheric Monitoring Network Salamanca (REDMAS - for its spanish acronym). The proposed models consider the use of meteorological variables as factors influencing the concentration of pollutants. The information used along this work is the current real data from REDMAS. In the proposed model, Artificial Neural Networks (ANN) combined with clustering algorithms are used. The type of ANN used is the Multilayer Perceptron with a hidden layer, using separate structures for the prediction of each pollutant. The meteorological variables used for prediction were: Wind Direction (WD), wind speed (WS), Temperature (T) and relative humidity (RH). Clustering algorithms, K-means and Fuzzy C-means, are used to find relationships between air pollutants and weather variables under consideration, which are added as input of the RNA. Those relationships provide information to the ANN in order to obtain the prediction of the pollutants. The results of the model proposed in this work are compared with the results of a multivariate linear regression and multilayer perceptron neural network. The evaluation of the prediction is calculated with the mean absolute error, the root mean square error, the correlation coefficient and the index of agreement. The results show the importance of meteorological variables in the prediction of the concentration of the pollutants SO2 and PM10 in the city of Salamanca, Gto., Mexico. The results show that the proposed model perform better than multivariate linear regression and multilayer perceptron neural network. The models implemented for each monitoring booth have the ability to make predictions of air quality that can be used in a system of real-time forecasting and human health impact analysis. Among the main results of the development of this thesis we can cite: A model based on artificial neural network combined with clustering algorithms for prediction with a hour ahead of the concentration of each pollutant (SO2 and PM10) is proposed. A different model was designed for each pollutant and for each of the three monitoring booths of the REDMAS. A model to predict the average of pollutant concentration in the next 24 hours of pollutants SO2 and PM10 is proposed, based on artificial neural network combined with clustering algorithms. Model was designed for each booth of the REDMAS and each pollutant separately. Resumen La contaminación atmosférica es una amenaza aguda, constituye un fenómeno que tiene particular incidencia sobre la salud del hombre. Los cambios que se producen en la composición química de la atmósfera pueden cambiar el clima, producir lluvia ácida o destruir el ozono, fenómenos todos ellos de una gran importancia global. La Organización Mundial de la Salud (OMS) considera la contaminación atmosférica como una de las más importantes prioridades mundiales. Salamanca, Gto., México; ha sido catalogada como una de las ciudades más contaminadas en este país. La industria de la zona propició un importante desarrollo económico y un crecimiento acelerado de la población en la segunda mitad del siglo XX. Las afectaciones en el aire son graves y se han hecho importantes esfuerzos por medir las concentraciones de los contaminantes. Las principales fuentes de contaminación son fuentes fijas como industrias químicas y de generación eléctrica. Los contaminantes que se han registrado como preocupantes son el Bióxido de Azufre (SO2) y las Partículas Menores a 10 micrómetros (PM10). La predicción de las concentraciones de estos contaminantes puede ser una potente herramienta que permita tomar medidas preventivas como reducción de emisiones a la atmósfera y alertar a la población afectada. En la presente tesis doctoral se propone un modelo de predicción de concentraci ón de los contaminantes más críticos SO2 y PM10 para cada caseta de monitorización de la Red de Monitorización Atmosférica de Salamanca (REDMAS). Los modelos propuestos plantean el uso de las variables meteorol ógicas como factores que influyen en la concentración de los contaminantes. La información utilizada durante el desarrollo de este trabajo corresponde a datos reales obtenidos de la REDMAS. En el Modelo Propuesto (MP) se aplican Redes Neuronales Artificiales (RNA) combinadas con algoritmos de agrupamiento. La RNA utilizada es el Perceptrón Multicapa con una capa oculta, utilizando estructuras independientes para la predicción de cada contaminante. Las variables meteorológicas disponibles para realizar la predicción fueron: Dirección de Viento (DV), Velocidad de Viento (VV), Temperatura (T) y Humedad Relativa (HR). Los algoritmos de agrupamiento K-means y Fuzzy C-means son utilizados para encontrar relaciones existentes entre los contaminantes atmosféricos en estudio y las variables meteorológicas. Dichas relaciones aportan información a las RNA para obtener la predicción de los contaminantes, la cual es agregada como entrada de las RNA. Los resultados del modelo propuesto en este trabajo son comparados con los resultados de una Regresión Lineal Multivariable (RLM) y un Perceptrón Multicapa (MLP). La evaluación de la predicción se realiza con el Error Medio Absoluto, la Raíz del Error Cuadrático Medio, el coeficiente de correlación y el índice de acuerdo. Los resultados obtenidos muestran la importancia de las variables meteorológicas en la predicción de la concentración de los contaminantes SO2 y PM10 en la ciudad de Salamanca, Gto., México. Los resultados muestran que el MP predice mejor la concentración de los contaminantes SO2 y PM10 que los modelos RLM y MLP. Los modelos implementados para cada caseta de monitorizaci ón tienen la capacidad para realizar predicciones de calidad del aire, estos modelos pueden ser implementados en un sistema que permita realizar la predicción en tiempo real y analizar el impacto en la salud de la población. Entre los principales resultados obtenidos del desarrollo de esta tesis podemos citar: Se propone un modelo basado en una red neuronal artificial combinado con algoritmos de agrupamiento para la predicción con una hora de anticipaci ón de la concentración de cada contaminante (SO2 y PM10). Se diseñó un modelo diferente para cada contaminante y para cada una de las tres casetas de monitorización de la REDMAS. Se propone un modelo de predicción del promedio de la concentración de las próximas 24 horas de los contaminantes SO2 y PM10, basado en una red neuronal artificial combinado con algoritmos de agrupamiento. Se diseñó un modelo para cada caseta de monitorización de la REDMAS y para cada contaminante por separado.

Hydrologic models for emergency decision support using Bayesian networks

In the presence of a river flood, operators in charge of control must take decisions based on imperfect and incomplete sources of information (e.g., data provided by a limited number sensors) and partial knowledge about the structure and behavior of the river basin. This is a case of reasoning about a complex dynamic system with uncertainty and real-time constraints where bayesian networks can be used to provide an effective support. In this paper we describe a solution with spatio-temporal bayesian networks to be used in a context of emergencies produced by river floods. In the paper we describe first a set of types of causal relations for hydrologic processes with spatial and temporal references to represent the dynamics of the river basin. Then we describe how this was included in a computer system called SAIDA to provide assistance to operators in charge of control in a river basin. Finally the paper shows experimental results about the performance of the model.

BaTboT: a biologically inspired flapping and morphing bat robot actuated by SMA-based artificial muscles

Bats are animals that posses high maneuvering capabilities. Their wings contain dozens of articulations that allow the animal to perform aggressive maneuvers by means of controlling the wing shape during flight (morphing-wings). There is no other flying creature in nature with this level of wing dexterity and there is biological evidence that the inertial forces produced by the wings have a key role in the attitude movements of the animal. This can inspire the design of highly articulated morphing-wing micro air vehicles (not necessarily bat-like) with a significant wing-to-body mass ratio. This thesis presents the development of a novel bat-like micro air vehicle (BaTboT) inspired by the morphing-wing mechanism of bats. BaTboT’s morphology is alike in proportion compared to its biological counterpart Cynopterus brachyotis, which provides the biological foundations for developing accurate mathematical models and methods that allow for mimicking bat flight. In nature bats can achieve an amazing level of maneuverability by combining flapping and morphing wingstrokes. Attempting to reproduce the biological wing actuation system that provides that kind of motion using an artificial counterpart requires the analysis of alternative actuation technologies more likely muscle fiber arrays instead of standard servomotor actuators. Thus, NiTinol Shape Memory Alloys (SMAs) acting as artificial biceps and triceps muscles are used for mimicking the morphing wing mechanism of the bat flight apparatus. This antagonistic configuration of SMA-muscles response to an electrical heating power signal to operate. This heating power is regulated by a proper controller that allows for accurate and fast SMA actuation. Morphing-wings will enable to change wings geometry with the unique purpose of enhancing aerodynamics performance. During the downstroke phase of the wingbeat motion both wings are fully extended aimed at increasing the area surface to properly generate lift forces. Contrary during the upstroke phase of the wingbeat motion both wings are retracted to minimize the area and thus reducing drag forces. Morphing-wings do not only improve on aerodynamics but also on the inertial forces that are key to maneuver. Thus, a modeling framework is introduced for analyzing how BaTboT should maneuver by means of changing wing morphology. This allows the definition of requirements for achieving forward and turning flight according to the kinematics of the wing modulation. Motivated by the biological fact about the influence of wing inertia on the production of body accelerations, an attitude controller is proposed. The attitude control law incorporates wing inertia information to produce desired roll (φ) and pitch (θ) acceleration commands. This novel flight control approach is aimed at incrementing net body forces (Fnet) that generate propulsion. Mimicking the way how bats take advantage of inertial and aerodynamical forces produced by the wings in order to both increase lift and maneuver is a promising way to design more efficient flapping/morphing wings MAVs. The novel wing modulation strategy and attitude control methodology proposed in this thesis provide a totally new way of controlling flying robots, that eliminates the need of appendices such as flaps and rudders, and would allow performing more efficient maneuvers, especially useful in confined spaces. As a whole, the BaTboT project consists of five major stages of development: - Study and analysis of biological bat flight data reported in specialized literature aimed at defining design and control criteria. - Formulation of mathematical models for: i) wing kinematics, ii) dynamics, iii) aerodynamics, and iv) SMA muscle-like actuation. It is aimed at modeling the effects of modulating wing inertia into the production of net body forces for maneuvering. - Bio-inspired design and fabrication of: i) skeletal structure of wings and body, ii) SMA muscle-like mechanisms, iii) the wing-membrane, and iv) electronics onboard. It is aimed at developing the bat-like platform (BaTboT) that allows for testing the methods proposed. - The flight controller: i) control of SMA-muscles (morphing-wing modulation) and ii) flight control (attitude regulation). It is aimed at formulating the proper control methods that allow for the proper modulation of BaTboT’s wings. - Experiments: it is aimed at quantifying the effects of properly wing modulation into aerodynamics and inertial production for maneuvering. It is also aimed at demonstrating and validating the hypothesis of improving flight efficiency thanks to the novel control methods presented in this thesis. This thesis introduces the challenges and methods to address these stages. Windtunnel experiments will be oriented to discuss and demonstrate how the wings can considerably affect the dynamics/aerodynamics of flight and how to take advantage of wing inertia modulation that the morphing-wings enable to properly change wings’ geometry during flapping. Resumen: Los murciélagos son mamíferos con una alta capacidad de maniobra. Sus alas están conformadas por docenas de articulaciones que permiten al animal maniobrar gracias al cambio geométrico de las alas durante el vuelo. Esta característica es conocida como (alas mórficas). En la naturaleza, no existe ningún especimen volador con semejante grado de dexteridad de vuelo, y se ha demostrado, que las fuerzas inerciales producidas por el batir de las alas juega un papel fundamental en los movimientos que orientan al animal en vuelo. Estas características pueden inspirar el diseño de un micro vehículo aéreo compuesto por alas mórficas con redundantes grados de libertad, y cuya proporción entre la masa de sus alas y el cuerpo del robot sea significativa. Esta tesis doctoral presenta el desarrollo de un novedoso robot aéreo inspirado en el mecanismo de ala mórfica de los murciélagos. El robot, llamado BaTboT, ha sido diseñado con parámetros morfológicos muy similares a los descritos por su símil biológico Cynopterus brachyotis. El estudio biológico de este especimen ha permitido la definición de criterios de diseño y modelos matemáticos que representan el comportamiento del robot, con el objetivo de imitar lo mejor posible la biomecánica de vuelo de los murciélagos. La biomecánica de vuelo está definida por dos tipos de movimiento de las alas: aleteo y cambio de forma. Intentar imitar como los murciélagos cambian la forma de sus alas con un prototipo artificial, requiere el análisis de métodos alternativos de actuación que se asemejen a la biomecánica de los músculos que actúan las alas, y evitar el uso de sistemas convencionales de actuación como servomotores ó motores DC. En este sentido, las aleaciones con memoria de forma, ó por sus siglas en inglés (SMA), las cuales son fibras de NiTinol que se contraen y expanden ante estímulos térmicos, han sido usados en este proyecto como músculos artificiales que actúan como bíceps y tríceps de las alas, proporcionando la funcionalidad de ala mórfica previamente descrita. De esta manera, los músculos de SMA son mecánicamente posicionados en una configuración antagonista que permite la rotación de las articulaciones del robot. Los actuadores son accionados mediante una señal de potencia la cual es regulada por un sistema de control encargado que los músculos de SMA respondan con la precisión y velocidad deseada. Este sistema de control mórfico de las alas permitirá al robot cambiar la forma de las mismas con el único propósito de mejorar el desempeño aerodinámico. Durante la fase de bajada del aleteo, las alas deben estar extendidas para incrementar la producción de fuerzas de sustentación. Al contrario, durante el ciclo de subida del aleteo, las alas deben contraerse para minimizar el área y reducir las fuerzas de fricción aerodinámica. El control de alas mórficas no solo mejora el desempeño aerodinámico, también impacta la generación de fuerzas inerciales las cuales son esenciales para maniobrar durante el vuelo. Con el objetivo de analizar como el cambio de geometría de las alas influye en la definición de maniobras y su efecto en la producción de fuerzas netas, simulaciones y experimentos han sido llevados a cabo para medir cómo distintos patrones de modulación de las alas influyen en la producción de aceleraciones lineales y angulares. Gracias a estas mediciones, se propone un control de vuelo, ó control de actitud, el cual incorpora información inercial de las alas para la definición de referencias de aceleración angular. El objetivo de esta novedosa estrategia de control radica en el incremento de fuerzas netas para la adecuada generación de movimiento (Fnet). Imitar como los murciélagos ajustan sus alas con el propósito de incrementar las fuerzas de sustentación y mejorar la maniobra en vuelo es definitivamente un tópico de mucho interés para el diseño de robots aéros mas eficientes. La propuesta de control de vuelo definida en este trabajo de investigación podría dar paso a una nueva forma de control de vuelo de robots aéreos que no necesitan del uso de partes mecánicas tales como alerones, etc. Este control también permitiría el desarrollo de vehículos con mayor capacidad de maniobra. El desarrollo de esta investigación se centra en cinco etapas: - Estudiar y analizar el vuelo de los murciélagos con el propósito de definir criterios de diseño y control. - Formular modelos matemáticos que describan la: i) cinemática de las alas, ii) dinámica, iii) aerodinámica, y iv) actuación usando SMA. Estos modelos permiten estimar la influencia de modular las alas en la producción de fuerzas netas. - Diseño y fabricación de BaTboT: i) estructura de las alas y el cuerpo, ii) mecanismo de actuación mórfico basado en SMA, iii) membrana de las alas, y iv) electrónica abordo. - Contro de vuelo compuesto por: i) control de la SMA (modulación de las alas) y ii) regulación de maniobra (actitud). - Experimentos: están enfocados en poder cuantificar cuales son los efectos que ejercen distintos perfiles de modulación del ala en el comportamiento aerodinámico e inercial. El objetivo es demostrar y validar la hipótesis planteada al inicio de esta investigación: mejorar eficiencia de vuelo gracias al novedoso control de orientación (actitud) propuesto en este trabajo. A lo largo del desarrollo de cada una de las cinco etapas, se irán presentando los retos, problemáticas y soluciones a abordar. Los experimentos son realizados utilizando un túnel de viento con la instrumentación necesaria para llevar a cabo las mediciones de desempeño respectivas. En los resultados se discutirá y demostrará que la inercia producida por las alas juega un papel considerable en el comportamiento dinámico y aerodinámico del sistema y como poder tomar ventaja de dicha característica para regular patrones de modulación de las alas que conduzcan a mejorar la eficiencia del robot en futuros vuelos.

Mutual Information and Perplexity based Clustering of Dialogue Information for Dynamic Adaptation of Language Models

We present two approaches to cluster dialogue-based information obtained by the speech understanding module and the dialogue manager of a spoken dialogue system. The purpose is to estimate a language model related to each cluster, and use them to dynamically modify the model of the speech recognizer at each dialogue turn. In the first approach we build the cluster tree using local decisions based on a Maximum Normalized Mutual Information criterion. In the second one we take global decisions, based on the optimization of the global perplexity of the combination of the cluster-related LMs. Our experiments show a relative reduction of the word error rate of 15.17%, which helps to improve the performance of the understanding and the dialogue manager modules.

Conductance interaction identification by means of Boltzmann distribution and mutual information analysis in conductance-based neuron models

Conductance interaction identification by means of Boltzmann distribution and mutual information analysis in conductance-based neuron models.