Memoria para optar al título de Ingeniera Civil Industrial / El dinamismo que ha presentado la industria del mejoramiento del hogar en el último tiempo, ha llevado a que las empresas involucradas deban preocuparse por entender el comportamiento de compra de sus consumidores, ya que no solo deben enfocar sus recursos y estrategias en capturar nuevos clientes sino también en la retención de éstos. El objetivo de este trabajo es estimar la fuga de clientes en una empresa de mejoramiento del hogar con el fin de generar estrategias de retención. Para ello se definirán criterios de fuga y se determinarán probabilidades para gestionar acciones sobre una fracción de clientes propensos a fugarse. Para alcanzar los objetivos mencionados, se trabajará sólo con clientes que forman parte de la cartera de un vendedor y se hará uso de las siguientes herramientas: estadística descriptiva, técnica RFM y la comparación de los modelos predictivos Árbol de decisión y Random Forest, donde la principal diferencia de estos últimos es la cantidad de variables y árboles que se construyen para la predicción de las probabilidades de fuga. Los resultados obtenidos entregan tres criterios de fuga, de manera que un cliente es catalogado como fugado cuando supera cualquiera de las cotas máximas, es decir, 180 días para el caso del recency, 20 para R/F o una variación de monto menores al -80%, por lo que la muestra queda definida con un 53,9% de clientes fugados versus un 46,1% de clientes activos. Con respecto a los modelos predictivos se tiene que el Árbol de decisión entrega un mejor nivel de certeza con un 84,1% versus un 74,7% del Random Forest, por lo que se eligió el primero obteniendo a través de las probabilidades de fuga 4 tipos de clientes: Leales (37,9%), Normales (7,8%), Propensos a fugarse (15,6%) y Fugados (38,7%). Se tiene que las causas de fuga corresponden a largos períodos de inactividad, atrasos en los ciclos de compras y una disminución en los montos y números de transacciones al igual que un aumento en el monto de transacciones negativas aludidas directamente a devoluciones y notas de crédito, por lo que las principales acciones de retención serían promociones, club de fidelización, descuentos personalizados y mejorar gestión en despachos y niveles de stock para que el cliente vuelva efectuar una compra en un menor plazo. Finalmente, a partir de este trabajo, se concluye que al retener 5% de clientes de probabilidades entre [0,5 y 0,75] y con el 50% de los mayores montos de transacciones se obtienen ingresos por USD $205 mil en 6 meses, representando el 5,5% de los clientes. Se propone validar este trabajo en nuevos clientes, generar alguna encuesta de satisfacción y mejorar el desempeño de los vendedores con una optimización de cartera.

Les vibrations aléatoires sont les sollicitations les plus représentatives à celles qui se trouvent dans l'environnement réel. Elles sont souvent rencontrées dans l'environnement de fonctionnement automobile, aéronautique, spatial, ferroviaire et d'autres. La particularité des vibrations aléatoires est l'utilisation de son analyse par fatigue. Grâce aux propriétés statistiques des signaux aléatoires que les méthodes spectrales permettent d'estimer la moyenne du dommage d'une façon efficace et économique mais avec certaines limitations. Une étude numérique sur un système électronique embarqué accompagnée des essais expérimentaux ont été mis en place dans cette thèse pour présenter une démarche complète d'une analyse spectrale des vibrations aléatoires. Une modélisation par éléments finis a été réalisée ainsi qu'une technique du 'zoom structural' en analyse spectrale des vibrations aléatoires a été proposée dans ce travail avec la présentation des équations de cette technique. Une application numérique a montré la validité de la technique proposée en appliquant une étude de l'endommagement par fatigue par l'approche temporelle et par l'approche spectrale. Les résultats numériques de cette application mènent à une estimation d'un temps à l'initiation de la fissure très proche du temps de défaillance observé dans les résultats des essais expérimentaux. Cette technique montre la performance et l'efficacité des méthodes spectrales dans l'estimation du dommage par fatigue aléatoire. Une étude de l'optimisation d'un système soumis aux sollicitations aléatoires a été développée. L'objectif de cette étude est de rechercher la conception optimale du système soumis aux vibrations aléatoires en posant des limitations sur le dommage qui en résulte. Ce dommage estimé par les méthodes spectrales de Dirlik et de Single Moment ne doit pas dépasser un dommage cible pour assurer la fiabilité du système étudié. / Random vibrations are the most representative excitations that can be found in the real environment. They are often encountered in the automotive, aeronautical, space, railway and other operating environments. The special feature of the random vibrations is their important role in damage fatigue analysis. The spectral methods allow to estimate efficiently and economically the mean of the damage using the statistical properties of the random signals. A numerical study on an embedded electronic system with experimental tests was set up in this thesis to present a complete spectral analysis of random vibrations methodology. A finite element model was performed as well as submodelling technique in spectral analysis of random vibrations with associated equations was proposed in this work. A numerical application has shown the validity of the proposed technique by applying the fatigue damage study using the time-domain approach and the spectral approach. The numerical results of this application lead to the estimation of a crack initiation time very close to the failure time observed in the experimental tests. This technique shows the performance and efficiency of spectral methods in the estimation of random fatigue damage. The optimization of a system subjected to random excitations has been developed in this work. The objective of this study is to retrieve the optimal design of the system subjected to random vibrations with limitations on the resulting damage. This damage estimated by Dirlik and Single Moment spectral methods must not exceed defined target damage in order to insure the reliability of the studied system.

Health plans are constantly seeking ways to assess and improve the quality of patient experience in various ambulatory and institutional settings. Standardized surveys are a common tool used to gather data about patient experience, and a useful measurement taken from these surveys is known as the Net Promoter Score (NPS). This score represents the extent to which a patient would, or would not, recommend his or her physician on a scale from 0 to 10, where 0 corresponds to "Extremely unlikely" and 10 to "Extremely likely". A large national health plan utilized automated calls to distribute such a survey to its members and was interested in understanding what factors contributed to a patient's satisfaction. Additionally, they were interested in whether or not NPS could be predicted using responses from other questions on the survey, along with demographic data. When the distribution of various predictors was compared between the less satisfied and highly satisfied members, there was significant overlap, indicating that not even the Bayes Classifier could successfully differentiate between these members. Moreover, the highly imbalanced proportion of NPS responses resulted in initial poor prediction accuracy. Thus, due to the non-linear structure of the data, and high number of categorical predictors, we have leveraged flexible methods, such as decision trees, bagging, and random forests, for modeling and prediction. We further altered the prediction step in the random forest algorithm in order to account for the imbalanced structure of the data.

A definir / A definir

Dans cette thèse nous explorons différentes facettes des modèles de tenseurs aléatoires. Les modèles de tenseurs aléatoires ont été introduits en physique dans le cadre de l'étude de la gravité quantique. En effet les modèles de matrices aléatoires, qui sont un cas particuliers de modèles de tenseurs, en sont une des origines. Ces modèles de matrices sont connus pour leur riche combinatoire et l'incroyable diversité de leurs propriétés qui les font toucher tous les domaines de l'analyse, la géométrie et des probabilités. De plus leur étude par les physiciens ont prouvé leur efficacité en ce qui concerne l'étude de la gravité quantique à deux dimensions. Les modèles de tenseurs aléatoires incarnent une généralisation possible des modèles de matrices. Comme leurs cousins, les modèles de matrices, ils posent questions dans les domaines de la combinatoire (comment traiter les cartes combinatoires d dimensionnelles ?), de la géométrie (comment contrôler la géométrie des triangulations générées ?) et de la physique (quel type d'espace-temps produisent-ils ? Quels sont leurs différentes phases ?). Cette thèse espère établir des pistes ainsi que des techniques d'études de ces modèles. Dans une première partie nous donnons une vue d'ensemble des modèles de matrices. Puis, nous discutons la combinatoire des triangulations en dimensions supérieures ou égales à trois en nous concentrant sur le cas tridimensionnelle (lequel est plus simple à visualiser). Nous définissons ces modèles et étudions certaines de leurs propriétés à l'aide de techniques combinatoires permettant de traiter les cartes d dimensionnelles. Enfin nous nous concentrons sur la généralisation de techniques issues des modèles de matrices dans le cas d'une famille particulières de modèles de tenseurs aléatoires. Ceci culmine avec le dernier chapitre de la thèse donnant des résultats partiels concernant la généralisation de la récurrence topologique de Eynard et Orantin à cette famille de modèles de tenseurs. / In this thesis manuscript we explore different facets of random tensor models. These models have been introduced to mimic the incredible successes of random matrix models in physics, mathematics and combinatorics. After giving a very short introduction to few aspects of random matrix models and recalling a physical motivation called Group Field Theory, we start exploring the world of random tensor models and its relation to geometry, quantum gravity and combinatorics. We first define these models in a natural way and discuss their geometry and combinatorics. After these first explorations we start generalizing random matrix methods to random tensors in order to describes the mathematical and physical properties of random tensor models, at least in some specific cases.

O avanço das pesquisas em redes complexas proporcionou desenvolvimentos significativos para a compreensão de sistemas complexos. Uma rede complexa é modelada matematicamente por meio de um grafo, onde cada vértice representa uma unidade dinâmica e suas interações são simbolizadas por um conjunto de arestas. Para se determinar propriedades estruturais desse sistema, caminhadas aleatórias tem-se mostrado muito úteis pois dependem apenas de informações locais (vértices vizinhos). Entre elas, destaca-se o passeio auto-excludente (SAW) que possui a restrição de não visitar um vértice que já foi alcançado, ou seja, apresenta memória do caminho percorrido. Por este motivo o SAW tem apresentado melhores resultados do que caminhantes sem restrição, na exploração da rede. Entretanto, por não se tratar de um processo Markoviano ele apresenta grande complexidade analítica, tornando indispensável o uso de simulações computacionais para melhor compreensão de sua dinâmica em diferentes topologias. Mesmo com as dificuldades analíticas, o SAW se tornou uma ferramenta promissora na identificação de estruturas de comunidades. Apesar de sua importância, detecção de comunidades permanece um problema em aberto devido à alta complexidade computacional associada ao problema de optimização, além da falta de uma definição formal do significado de comunidade. Neste trabalho, propomos um método de detecção de comunidades baseado em SAW para extrair uma estrutura de comunidades da rede otimizando o parâmetro modularidade. Combinamos características extraídas desta dinâmica com a análise de componentes principais para posteriormente classificar os vértices em grupos por meio da clusterização hierárquica aglomerativa. Para avaliar a performance deste novo algoritmo, comparamos os resultados com outras quatro técnicas populares: Girvan-Newman, Fastgreedy, Walktrap e Infomap, aplicados em dois tipos de redes sintéticas e nove redes reais diversificadas e bem conhecidas. Para os benchmarks, esta nova técnica produziu resultados satisfatórios em diferentes combinações de parâmetros, como tamanho de rede, distribuição de grau e número de comunidades. Já para as redes reais, obtivemos valores de modularidade superior aos métodos tradicionais, indicando uma distribuição de grupos mais adequada à realidade. Feito isso, generalizamos o algoritmo para redes ponderadas e digrafos, além de incorporar metadados à estrutura topológica a fim de melhorar a classificação em grupos. / The progress in complex networks research has provided significant understanding of complex systems. A complex network is mathematically modeled by a graph, where each vertex represents a dynamic unit and its interactions are symbolized by groups of edges. To determine the system structural properties, random walks have shown to be a useful tool since they depend only on local information (neighboring vertices). Among them, the selfavoiding walk (SAW) stands out for not visiting vertices that have already been reached, meaning it can record the path that has been travelled. For this reason, SAW has shown better results when compared to non-restricted walkers network exploration methods. However, as SAW is not a Markovian process, it has a great analytical complexity and needs computational simulations to improve its dynamics in different topologies. Even with the analytical complexity, SAW has become a promising tool to identify the community structure. Despite its significance, detecting communities remains an unsolved problem due to its high computational complexity associated to optimization issues and the lack of a formal definition of communities. In this work, we propose a method to identify communities based on SAW to extract community structure of a network through optimization of the modularity score. Combining technical features of this dynamic with principal components analyses, we classify the vertices in groups by using hierarchical agglomerative clustering. To evaluate the performance of this new algorithm, we compare the results with four other popular techniques: Girvan-Newman, Fastgreedy, Walktrap and Infomap, applying the algorithm in two types of synthetic networks and nine different and well known real ones. For the benchmarks, this new technique shows satisfactory results for different combination of parameters as network size, degree distribution and number of communities. As for real networks, our data shows better modularity values when compared to traditional methods, indicating a group distribution most suitable to reality. Furthermore, the algorithm was adapted for general weighted networks and digraphs in addition to metadata incorporated to topological structure, in order to improve the results of groups classifications.

Campos Aleatórios Condicionais são modelos probabilísticos discriminativos que tem sido utilizados com sucesso em diversas áreas como processamento de linguagem natural, reconhecimento de fala e bioinformática. Entretanto, implementar algoritmos eficientes para esse tipo de modelo não é uma tarefa fácil. Nesse trabalho apresentamos um arcabouço que ajuda no desenvolvimento e experimentação de Campos Aleatórios Condicionais Semi Markovianos (semi-CRFs). Desenvolvemos algoritmos eficientes que foram implementados em C++ propondo uma interface de programação flexível e intuitiva que habilita o usuário a definir, treinar e avaliar modelos. Nossa implementação foi construída como uma extensão do arcabouço ToPS que, inclusive, pode utilizar qualquer modelo já definido no ToPS como uma função de característica especializada. Por fim utilizamos nossa implementação de semi-CRF para construir um preditor de promotores que apresentou performance superior aos preditores existentes. / Conditional Random Fields are discriminative probabilistic models that have been successfully used in several areas like natural language processing, speech recognition and bioinformatics. However, implementing efficient algorithms for this kind of model is not an easy task. In this thesis we show a framework that helps the development and experimentation of Semi-Markov Conditional Random Fields (semi-CRFs). It has an efficient implementation in C++ and an intuitive API that allow users to define, train and evaluate models. It was built as an extension of ToPS framework and can use ToPS probabilistic models as specialized feature functions. We also use our implementation of semi-CRFs to build a high performance promoter predictor.

Function words are known to be shorter than content words. I investigate the function words BE and HAVE (with its content word homonym) and show that more reduction, operationalized as word shortening or contraction, is found in some grammaticalized meanings of these words. The difference between the words’ uses cannot be attributed to differences in frequency or semantic weight. Instead I argue that these words are often shortened and reduced when they occur in constructions in which they are highly predictable. This suggests that particular grammaticalized uses of a word are stored with their own exemplar clouds of context-specific phonetic realizations. The phonetics of any instance of a word are then jointly determined by the exemplar cloud for that word and the particular context. A given instance of an auxiliary can be reduced either because it is predictable in the current context or because that use of the auxiliary usually occurs in predictable contexts. The effects cannot be attributed to frequency or semantic weight. The present study compares function word production in the speech of school-aged children and their caregivers and in inter-adult speech. The effects of predictability in context and average predictability across contexts are replicated across the datasets. However, I find that as children get older their function words shorten relative to content words, even when controlling for increasing speech rate, showing that as their language experience increases they spend less time where it is not needed for comprehensibility. Caregivers spend less time on function words with older children than younger children, suggesting that they expect function words to be more difficult for younger interlocutors to decode than for older interlocutors. Additionally, while adults use either word shortening or contraction to increase the efficiency of speech, children tend to either use contraction and word shortening or neither until age seven, where they start to use one strategy or the other like adults. Young children with better vocabulary employ an adult-like strategy earlier, suggesting earlier onset of efficient yet effective speech behavior, namely allocating less signal to function words when they are especially easy for the listener to decode.

Num problema de espalhamento temos partículas incidentes sobre uma região de espalhamento que, depois de interagir por algum tempo nessa região, escapam para o infinito. Quando o espalhamento é caótico, a função de espalhamento (que é a relação entre uma variável antes do espalhamento e outra variável depois do espalhamento), apresenta singularidades sobre um conjunto de Cantor de condições iniciais. O espalhamento caótico pode ser dividido em dois tipos: espalhamento não-hiperbólico e hiperbólico. No espalhamento não-hiperbólico, o conjunto invariante contém órbitas estáveis. O decaimento das partículas que escapam do conjunto invariante é regido por uma lei de potência com relação ao tempo. No caso do espalhamento hiperbólico, a sela caótica é hiperbólica e todas as órbitas que a compõem são instáveis. O decaimento das partículas na região de espalhamento segue uma exponencial decrescente. Investigamos a transição do espalhamento não-hiperbólico para o hiperbólico quando ruído é adicionado à dinâmica do sistema. Isto porque prevíamos que o ruído reduzisse o efeito de aprisionamento (stickness) dos conjuntos de órbitas estáveis, provocando um decaimento exponencial. Introduzimos perturbações randômicas a fim de simular flutuações reais que ocorrem em sistemas físicos, como por exemplo, um vórtex que depende irregularmente do tempo no estudo de fluidos. Assim, usamos o conceito de mapas randômicos, que são mapas onde um ou mais parâmetros são variados aleatoriamente a cada iteração. Estudamos então, os efeitos provocados por perturbações randômicas em um sistema com espalhamento caótico não-hiperbólico. / In a scattering problem we have particles inciding on a scattering region and these particles, after spending some time in this region, escape towards infinity. When the scattering is chaotic, the scattering function (a function that relates an input variable with an output variable), is singular on a Cantor set of initial conditions. The chaotic scattering can be either non-hyperbolic or hyperbolic. In the non-hyperbolic scattering, the invariant set has stable orbits. This decay is governed by a power law in time. In the hyperbolic case, the chaotic saddle is hyperbolic and all the orbits are unstable. The decay of the particles is a decreasing exponential in the time. We investigate the transition from non-hyperbolic to hyperbolic scattering as noise is added to the system. One expects that noise will reduce the stickness of the regular regions, resulting in an exponential decay law, typical of hyperbolic systems. We apply random perturbations in order to simulate the real fluctuations that occur in physical systems, for example, an aperiodic vortex in a fluid flow. So, we work with random maps, where we change randomly one or more parameters on each iteration. We study thus, the effects of the random perturbations on a system having non-hyperbolic scattering.

by Chi-van Lam. / Thesis (M.Phil.)--Chinese University of Hong Kong, 1993. / Includes bibliographical references (leaves 71-72). / Chapter Chapter 1 --- Introduction --- p.1 / Chapter Chapter 2 --- The Model and the Bayes risk --- p.7 / Chapter § 2.1 --- The Model / Chapter § 2.2 --- The Bayes risk / Chapter Chapter 3 --- The Algorithm --- p.16 / Chapter § 3.1 --- A finite algorithm / Chapter § 3.2 --- The Number Theoretical Method for Optimization / Chapter § 3.2.1 --- NTMO / Chapter § 3.2.2 --- SNTMO / Chapter Chapter 4 --- Quadratic Loss Function --- p.26 / Chapter §4.1 --- The Bayes risk / Chapter § 4.2 --- An optimal plan / Chapter § 4.3 --- Numerical Examples / Chapter Chapter 5 --- Conclusions and Comments --- p.42 / Chapter § 5.1 --- Comparison between various plans / Chapter § 5.2 --- Sensitivity Analysis / Chapter § 5.3 --- Further Developments / Tables --- p.46 / Appendix A --- p.60 / Appendix B --- p.65 / References --- p.71