Dans cette thèse, nous donnons une localisation en espace de la théorie des processus de Feller. Un premier objectif est d’obtenir des résultats simples et précis sur la convergence de processus de Markov. Un second objectif est d’étudier le lien entre les notions de propriété de Feller, problème de martingales et topologie de Skorokhod. Dans un premier temps nous donnons une version localisée de la topologie de Skorokhod. Nous en étudions les notions de compacité et tension. Nous faisons le lien entre les topologies de Skorokhod localisée et non localisée, grâce à la notion de changement de temps. Dans un second temps, à l’aide de la topologie de Skorokhod localisée et du changement de temps, nous étudions les problèmes de martingales. Nous montrons pour des processus l’équivalence entre, d’une part, être solution d’un problème de martingales bien posé, d’autre part, vérifier une version localisée de la propriété de Feller, et enfin, être markovien et continu en loi par rapport à sa condition initiale. Nous caractérisons la convergence en loi pour les solutions de problèmes de martingale en terme de convergence des opérateurs associés et donnons un résultat similaire pour les approximations à temps discret. Pour finir, nous appliquons la théorie des processus localement fellerien à deux exemples. Nous l’appliquons d’abord au processus de type Lévy et obtenons des résultats de convergence pour des processus à temps discret et continu, notamment des méthodes de simulation et schémas d’Euler. Nous appliquons ensuite cette même théorie aux diffusions unidimensionnelles dans des potentiels, nous obtenons des résultats de convergence de diffusions ou marches aléatoires vers des diffusions singulières. Comme conséquences, nous déduisons la convergence de marches aléatoires en milieux aléatoires vers des diffusions en potentiels aléatoires. / In this PhD thesis, we give a space localisation for the theory of Feller processes. A first objective is to obtain simple and precise results on the convergence of Markov processes. A second objective is to study the link between the notions of Feller property, martingale problem and Skorokhod topology. First we give a localised version of the Skorokhod topology. We study the notions of compactness and tightness for this topology. We make the connexion between localised and unlocalised Skorokhod topologies, by using the notion of time change. In a second step, using the localised Skorokhod topology and the time change, we study martingale problems. We show the equivalence between, on the one hand, to be solution of a well-posed martingale problem, on the other hand, to satisfy a localised version of the Feller property, and finally, to be a Markov process weakly continuous with respect to the initial condition. We characterise the weak convergence for solutions of martingale problems in terms of convergence of associated operators and give a similar result for discrete time approximations. Finally, we apply the theory of locally Feller process to some examples. We first apply it to the Lévy-type processes and obtain convergence results for discrete and continuous time processes, including simulation methods and Euler’s schemes. We then apply the same theory to one-dimensional diffusions in a potential and we obtain convergence results of diffusions or random walks towards singular diffusions. As a consequences, we deduce the convergence of random walks in random environment towards diffusions in random potential.

Le travail développe un modèle de portfolio à propos de la dollarisation financière (FD), et l'estime pour la Russie. La contribution de ce travail sera de construire le premier modèle théorique de variance moyenne asymétrique d'aplatissement sur la dollarisation financière et de le valider empiriquement. Le travail se fonde sur des recherches antérieures qui ont trouvé que l'ajout de moments plus élevés, comme l'asymétrie et l'aplatissement, à la variance minimale du portfolio(MVP) permettant une meilleure modélisation des choix de portfolio et de développe un model comme celui-ci pour la FD. Nous utilisons ensuite les méthodes Markovswitching sur les données mensuelles pour les dépôts bancaires en Russie depuis la fin des années 1990 afin de documenter l'influence dominante de l'inflation et de la dépréciation de la monnaie et de leurs moments comme principaux déterminants de dépôt de dollarisation dans un cadre de variance-moyenne-asymétrique-aplatie en période de crise, par opposition aux périodes normales. / This thesis develops a portfolio model of financial dollarization (FD) and estimates it for Russia. The contribution of this work will be to construct the first theoretical meanvariance-skewness-kurtosis model of financial dollarization and to validate it empirically. The work builds on previous research which found that adding higher moments, as Skewness and Kurtosis, to the minimum variance portfolio (MVP) enables a better modelling of portfolio choice, and develops such a model for FD. We then use Markovswitching methods on monthly data for bank deposits in Russia since the late 1990s to document the dominant influence of inflation and currency depreciation and their moments as the main determinants of deposit dollarization in a mean-varianceskewness-kurtosis framework during crisis as opposed to normal periods.

A Markov chain model for the early stochastic phase of the transmission of an infectious pathogen is studied, investigating its properties in the case of an isolated population and of two coupled populations with explicit movement of infectious individuals. Travel was found to play a role in the early development and spread of an infectious disease, particularly in the case of differing basic reproduction numbers in the connected locations. / February 2017

A Dissertation submitted to the Faculty of Science, University of the Witwatersrand, Johannesburg, in ful lment of the requirements for the degree of Master of Science. Johannesburg, 2016. / Hidden Markov models (HMMs) and double chain Markov models (DCMMs) are classical Markov model extensions used in a range of applications in the literature. This dissertation provides a comprehensive review of these models with focus on i) providing detailed mathematical derivations of key results - some of which, at the time of writing, were not found elsewhere in the literature, ii) discussing estimation techniques for unknown model parameters and the hidden state sequence, and iii) discussing considerations which practitioners of these models would typically take into account. Simulation studies are performed to measure statistical properties of estimated model parameters and the estimated hidden state path - derived using the Baum-Welch algorithm (BWA) and the Viterbi Algorithm (VA) respectively. The effectiveness of the BWA and the VA is also compared between the HMM and DCMM. Selected HMM and DCMM applications are reviewed and assessed in light of the conclusions drawn from the simulation study. Attention is given to application in the field of Credit Risk. / LG2017

Thesis (M.A.)--Boston University / Probability problems in which a time parameter is involved are known as stochastic processes. The simplest time dependent stochastic processes are those in which the probabilities of a system changing to various states are solely dependent upon the present state of the system. These processes are known as Markov processes, or for the case where only discrete time intervals are considered, as Markov chains. A Markov chain may be completely defined by the matrix of its transition probabilities. This matrix is called a stochastic matrix and is characterized by the facts that it is a square matrix, that the elements of each column sum to one and that all the elements are non-negative. An important consideration in most Markov chain problems is the effect of a number of transitions as defined by the stochastic matrix. Performing this operation requires determining the higher powers of the stochastic matrix. Two modal matrices are defined, where k is the matrix of the column characteristic vectors of the stochastic matrix and K is the matrix of the row characteristic vectors. It is shown that with proper normalization of these vectors, the stochastic matrix P is equal to kAK, where A is the matrix of the characteristic roots along the diagonal and zeroes elsewhere. .The higher powers of the stochastic matrix, Pm, are then found to be equal to kAmk. The stochastic matrix is found always to have a characteristic root one, and all the other roots are shown to be less than one in absolute value. The limiting transition matrix P ∞ is found to have identical columns, each consisting of the characteristic column vector associated with the characteristic root one. The limiting distribution is the same vector and is independent of the initial conditions.[TRUNCATED]

Thèse de doctorat : Informatique : Nancy 1 : 2008. / Titre provenant de l'écran-titre.

Reproduction de : Thèse de doctorat : mathématiques appliquées : Clermont-Ferrand 2 : 2005. / Thèse bilingue. Titre provenant de l'écran-titre. Bibliographie dispersée.

Thesis (Ph. D.)--University of Washington, 2001. / Vita. Includes bibliographical references (leaves 53-54).

Thesis (M. Phil.)--University of Hong Kong, 1989.

In this thesis, the properties of some non-standard Markov chain models and their corresponding parameter estimation methods are investigated. Several practical applications and extensions are also discussed. The estimation of model parameters plays a key role in the real-world applications of Markov chain models. Some widely used estimation methods for Markov chain models are based on the existence of stationary vectors. In this thesis, some weaker sufficient conditions for the existence of stationary vectors for highorder Markov chain models, multivariate Markov chain models and high-order multivariate Markov chain models are proposed. Furthermore, for multivariate Markov chain models, a new estimation method based on minimizing the prediction error is proposed. Numerical experiments are conducted to demonstrate the efficiency of the proposed estimation methods with an application in demand prediction. Hidden Markov Model (HMM) is a bivariate stochastic process such that one of the process is hidden and the other is observable. The distribution of observable sequence depends on the hidden sequence. In a traditional HMM, the hidden states directly affect the observable states but not vice versa. However, in reality, observable sequence may also have effect on the hidden sequence. For this reason, the concept of Interactive Hidden Markov Model (IHMM) is introduced, whose key idea is that the transitions of the hidden states depend on the observable states too. In this thesis, efforts are devoted in building a highorder IHMM where the probability laws governing both observable and hidden states can be written as a pair of high-order stochastic difference equations. We also propose a new model by capturing the effect of observable sequence on the hidden sequence through using the threshold principle. In this case, reference probability methods are adopted in estimating the optimal model parameters, while for unknown threshold parameter, Akaike Information Criterion (AIC) is used. We explore asset allocation problems from both domestic and foreign perspective where asset price dynamics follows autoregressive HMM. The object of an investor is not only to maximize the expected utility of the terminal wealth, but also to ensure that the risk of the portfolio described by the Value-at-Risk (VaR) does not exceed a specified level. In many decision processes, fuzziness is a major source of imprecision. As a perception of usual Markov chains, the definition of fuzzy Markov chains is introduced. Compared to traditional Markov chain models, fuzzy Markov chains are relatively new and many properties of them are still unknown. Due to the potential applications of fuzzy Markov chains, we provide some characterizations to ensure the ergodicity of these chains under both max-min and max-product compositions. / published_or_final_version / Mathematics / Doctoral / Doctor of Philosophy