資產證券化一詞源自1970年代，而第一筆資產基礎證券的發行則始於1986 年。由於其具有自資本市場直接融資、降低籌資成本、分散籌資來源、提高資本適足率等誘因之下，逐漸受到銀行以及其他企業的重視，並發展成為固定收益證券市場中比重相當大的一環。擔保債權證券交易自1988年出現在美國，然後在歐美迅速發展，目前已成為重要債券市場之一。台灣金融產業發展正值轉型期，銀行除面對低利率帶來經營壓力之外，同時亦需規避評等較差之企業貸款的信用風險，而保險業者在低利率時代來臨卻無良好報酬之投資標的可供投資。因此，此環境乃為推動證券化之良好契機。自1997年發生東南亞金融危機，乃至1998年韓國的亞洲金融危機，造成許多跨國企業紛紛裁員、關廠、甚至倒閉，造成一連串的金融危機連鎖效應。因此，公司間或產業間之榮枯是相互關聯的，且均會受總體經濟因素所影響。是以，近年來信用風險亦成為近年來財務領域上重要議題。理論或實證上，當多個標的資產之信用衍生性商品被加以開發，並用來管理信用風險的時候，需考慮多個標的資產間的違約相關性，方能準確地衡量信用風險。故在信用風險管理與信用衍生性商品的評價中，違約相關性的估計與考量顯得格外重要。結構式或縮減式模型在發展違約相關性的多變數模型中是困難的，因為其衍生性商品價值的理論推導繁複或其數值計算是相當費時。本研究是假設隨機違約強度模式下，並在多標的資產之信用風險評價模型中，透過適當個別資產之邊際違約機率與Copula函數之選擇，及其相關參數之估算，即可快速求算具違約相關性之多變數聯合機率函數，以利擔保債權證券(CDO)之評價，並模擬出未來可能損失分配，進行主次順位架構下不同分券比例的敏感度分析，以瞭解不同的次順位劃分比例對於各順位分券風險值的影響程度為何。另外，隨著國內目前證券化腳步的發展，在未來證券化商品勢必成為市場上的主流商品之一，將來可供證券化的資產種類勢必也會增加。因此，為了提高投資人對於不同資產種類的投資信心，以方便發行人對於證券化商品的銷售，信用增強機制在證券化中所佔的地位也將更形重要。在這樣的情況下，評估標的資產可能的違約損失，以決定信用增強的比例該為多少，對於發行人而言也將會是一項重要的課題。因此本文針對Copula方法與分析架構做一剖析，再以國內第一檔公募之擔保債權證券－法國里昂信貸銀行企業貸款為例，進行模擬實證並分析結果。 本研究結論如下：在信用風險管理與信用衍生性商品評價中，違約相關性是ㄧ個重要的因子。此外，本研究發現違約回收率、標的資產間的相關係數以及違約機率等三者均會影響分券信用價差的評價：就權益分券而言，信用價差與相關係數是呈反比的，而次償分券表現出來也與權益分券大致相同，相對於權益分券與次償分券，先償分券之信用價差則與債權群組內標的資產間的相關係數則是呈正比的。實證結果亦顯示，承購風險性較高的分券其估算出的合理的風險溢酬也較高，此外，當違約回收率愈高時，債務人違約後的損失愈低，因此發行者需給予證券投資人的合理風險溢酬也愈低。另外，債務人的信用評等高低影響違約機率的大小，其亦是影響CDO商品合理溢酬高低的主要關鍵。再者，假使忽略債務人之間的違約相關性時，則各分券所估算出來的合理溢酬均會有所偏誤。因此，此結果隱含，評價CDO商品之合理溢酬時，需考慮債權群組內資產間的違約相關性，亦即投資人所面對風險除分券本身的信用風險外，還需考慮到債權人違約相關之風險，違約相關性愈高則投資者所面對的風險也較高。另外，介於先償分券與權益分券之間的次償債之信用價差，其信用價差大小大致上約介於先償分券與權益分券之間，而預期損失金額則是受到各順位分券比例大小不同，而有所差異。因此，將再進行主次順位架構下不同分券比例的敏感度分析，結果顯示若次順位債權分券（即權益分券）比例愈高，愈有信用增強之作用。 最後，若將本研究所估算出來分券的合理溢酬與實際CDO契約所載明計算出的結果相比較，其間差異並不大。因此本研究所建構的評價模型應能提供發行者與市場投資人一個評價基礎，不失為一種可行方法。

在國際間的股票市場中，股票報酬常存在有不對稱的相關結構，而其會造成許多極度地尾端風險。Copula函數常被用來描述多變數之間的聯合相關程度。多數的文獻均以二元copula函數為架構，去描述多種不同資產，像是股票、債券、匯率等之間的關係。我們討論多元copula的應用，本文以四元copula為主軸，並輔以狀態轉換 (regime-switching) 之機率過程，建構出四資產的投資組合之相關結構模型。 考慮了狀態轉換之copula的配適性後，我們以此模型來做資產投資策略。在模擬過程中，我們嘗試根據不同的未來目標做出最佳的投資組合權重，並採用動態預期模型 (dynamic anticipative model) 來藉由資訊的不斷更新，重新估計模型的參數來做資產評估。實證結果上，我們發現考慮狀態轉換之copula模型可以捕捉到更多股票報酬波動的情形，因此能減少在股市共跌時造成的重大損失。 / The correlation of returns in international stock markets exist asymmetric structure, which cause extremely tail dependence. The copula functions are commonly used to describe the dependence between random variables. Most literatures use basic pair-copulas to model the dependence of two variables, like stocks, bonds and exchange rates. This article try to use multivariate copulas, mainly 4-copula, and regime-switching method to construct a portfolio dependence, and extend to asset allocation. Given the fitting regime-switching copula, we use the model to decide investment strategy. We try to select the optimal weights of portfolio by different objective function, and we adapt a dynamic anticipative model, which can take all new information for parameters estimation. Empirically, we find that the copula-based model with regime-switching can capture more variation, and decrease the return loss from downside co-movement.

In this thesis, I explore the use of copulas in non-copula constructions in the languages of South Asia to establish possible genetic and areal tendencies in the distribution. Using materials – language descriptions and data – from Grierson’s Linguistic Survey of India, I examine the phenomenon in 206 languages from four families (Munda, Dravidian, Indo-Aryan and Sino-Tibetan). It is found that the languages of South Asia appear to be more likely than the world-wide average to use the copula in non-copula constructions and that at least Munda, Dravidian and Indo-Aryan use it in the same way with regards to tense, namely in the past and present but not the future. Finally, I argue that there is some evidence supporting that the use of the copula in non-copula constructions is an areal feature, though more work is needed to make any definitive conclusions.

Os modelos de fragilidade são utilizados para modelar as possíveis associações entre os tempos de sobrevivência. Uma outra alternativa desenvolvida para modelar a dependência entre dados multivariados é o uso dos modelos baseados em funções cópulas. Neste trabalho propusemos dois modelos de sobrevivência derivados das cópulas de Ali- Mikhail-Haq (AMH) e de Frank para modelar a dependência de dados bivariados na presença de covariáveis e observações censuradas. Para fins inferenciais, realizamos uma abordagem bayesiana usando métodos Monte Carlo em Cadeias de Markov (MCMC). Algumas discussões sobre os critérios de seleção de modelos são apresentadas. Com o objetivo de detectar observações influentes utilizamos o método bayesiano de análise de influência de deleção de casos baseado na divergência ψ. Por fim, mostramos a aplicabilidade dos modelos propostos a conjuntos de dados simulados e reais. Apresentamos, também, um novo modelo de sobrevivência bivariado com fração de cura, que leva em consideração três configurações para o mecanismo de ativação latente: ativação aleatória, primeira ativação é última ativação. Aplicamos este modelo a um conjunto de dados de empréstimo de Crédito Direto ao modo do Consumidor (DCC) e comparamos os ajustes por meio dos critérios bayesianos de seleção de modelos para verificar qual dos três modelos melhor se ajustou. Por fim, mostramos nossa proposta futura para a continuação da pesquisa. / The frailty models are used to model the possible associations between survival times. Another alternative developed for modeling the dependence between multivariate data is the use of models based on copulas functions. In this paper we propose two derived survival models of copula of the Ali-Mikhail-Haq (AMH) and of the Frank to model the dependence of bivariate data in the presence of covariates and censored observations. For inferential purposes, we conducted a Bayesian approach using Monte Carlo methods in Markov Chain (MCMC). Some discussions on the model selection criteria were presented. In order to detect influential observations we use the Bayesian method of cases of deletion of influence analysis based on the difference ψ. Finally, we show the applicability of the proposed models to sets of simulated and real data. We present, too, a new survival model with bivariate fraction of healing, which takes into account three settings for the latent activation mechanism: random activation, first activation and final activation. We apply this model to a set of Direct Credit loan data to the Consumer mode (DCC) and compare the settings, through Bayesian criteria for selection of models, which of the three models best fit. Finally, we show our future proposal for further research.

Neste trabalho objetivou-se estudar o comportamento de acasalamento de Cyrtomon luridus Boheman (Coleoptera: Curculionidae), bem como a presença de semioquímicos mediando o acasalamento e a interação com a planta hospedeira Duboisia sp. (Solanaceae). Machos e fêmeas adultos recém-emergidos foram coletados no campo e levados para o laboratório. Incialmente foi determinado o melhor parâmetro morfológico para diferenciação de machos e fêmeas. Verificou-se que machos deste curculionídeo possuem a abertura genital oclusa pelo último tergito. O comportamento de acasalamento foi observado em casais virgens durante nove dias. Verificou-se que machos e fêmeas de C. luridus atingiram a maturidade sexual dois dias após a emergência. Machos e fêmeas realizaram acasalamentos repetidos com o mesmo parceiro em qualquer horário do dia. O comportamento de acasalamento deste curculionídeo foi dividido nas fases pré-copulatória, copulatória e pós-copulatória. Além disso, estudos de olfatometria revelaram que machos e fêmeas de C. luridus foram fortemente atraídos por voláteis de machos se alimentando sobre a planta hospedeira. A coleta e análise destes voláteis revelaram a presença dos compostos (Z)-3-hexenal, hexanal, (E)-2-hexenal, (E)-2-hexen-1-ol, fenilacetaldeido, linalol e geraniol, como eletrofisiologicamente ativos. O presente trabalho traz importantes informações sobre o comportamento de acasalamento e ecologia química de Entiminae e servirão de base para estudos com outras espécies desta importante subfamília de Curculionidae. / This work aimed to study the mating behavior of Cyrtomon luridus Boheman (Coleoptera: Curculionidae), as well as the presence of semiochemicals mediating the mating and the interaction of the insects with the host plant, Duboisia sp. (Solanaceae). Newly emerged males and females adults were collected in the field and taken to the laboratory. Initially it determined the best morphological parameter to differentiate males and females. It was found that males possess the genital opening occluded by the last tergite. The mating behavior was observed in virgin couples for nine days. It was found that males and females C. luridus reached sexual maturity two days after emergence. Males and females made repeated matings with the same partner at any time of day. The mating behavior was clearly divided into the pre-copulatory, copulatory, and post-copulatory phases. In addition, olfatometrics studies revealed that males and females C. luridus were strongly attracted to volatile males feeding on the host plant. The collection and analysis of these volatiles revealed the compounds (Z)-3-hexenal, hexanal, (E)-2-hexenal, (E)-2-hexen-1-ol, phenylacetaldehyde, linalool and geraniol as being eletrophysiologically. The present work brings important information about the mating behavior and chemical ecology of Entiminae, which will serve as support for studies with other species of this important subfamily of Curculionidae.

Orientador: Jorge Alberto Achcar / Tese (doutorado) - Universidade Estadual de Campinas, Instituto de Matemática, Estatística e Computação Científica / Made available in DSpace on 2018-08-18T02:57:38Z (GMT). No. of bitstreams: 1 TovarCuevas_JoseRafael_D.pdf: 2314079 bytes, checksum: 3de8e433bd4d02c785316e5d57f1a980 (MD5) Previous issue date: 2011 / Resumo: A maioria dos estudos sobre estimação da prevalência e parâmetros de desempenho de testes para diagnóstico clínico não tem considerado que muitos dos métodos de diagnóstico incluem a medição de traços biológicos cuja resposta é expressa em escala contínua e que, devido ao fato de serem medidos no mesmo indivíduo, esses traços necessariamente apresentam algum tipo de dependência que pode ou não ser explicada como um fenômeno de comportamento linear ou de concordância. Além disso, a análise de dados realizada nesses estudos parte do pressuposto de que a estrutura dos testes é binária sem considerar o fato de que as observações assumem essa apresentação depois de serem dicotomizadas usando um ponto de corte estabelecido a partir de critérios clínicos. Nesta tese, apresenta-se uma proposta de abordagem Bayesiana ao problema da estimação da prevalência, da sensibilidade e da especificidade dos testes dentro de planejamentos que incluem a aplicação de dois ou três testes diagnósticos de triagem, os quais são produto da medição de igual número de traços biológicos expressos em escala contínua com ponto de corte para dicotimização e um padrão-ouro para verificação. Embora o objetivo principal do modelo estatístico proposto seja estudar o efeito da dependência entre resultados dos testes de triagem sobre as estimativas da prevalência e os parâmetros de desempenho, também se consideram alternativas para contornar outras dificuldades comuns neste tipo de estudos, como a falta de identificabilidade e o viés devido à não verificação com padrão-ouro de indivíduos com resultado negativo em ambos os testes de triagem (viés de verificação). A proposta considera o uso de três funções cópula para modelar a dependência e a avaliação de três níveis da mesma. Dado o enfoque Bayesiano do estudo, foi necessário desenvolver um procedimento para elicitar distribuições a priori em situações de total ausência de informação sobre o parâmetro de interesse, o que acontece com as cópulas, funções bastante desconhecidas na pesquisa médica. Os resultados obtidos com o modelo proposto foram comparados com aqueles obtidos utilizando a covariância como parâmetro de dependência e o pressuposto de independência. O modelo apresenta uma reparametrização que, a diferença da maioria dos métodos apresentados na literatura sobre o tema, permite obter diretamente as estimativas de interesse sem a necessidade de complexos procedimentos analíticos e computacionais. A presença de dependência tem pouco efeito sobre a estimativa da prevalência e afeta as estimativas dos parâmetros de desempenho, o efeito é mais forte quando o planejamento apresenta viés de verificação. Dependências fracas subestimam as sensibilidades e os outros parâmetros não apresentam viés, enquanto que dependências fortes superestimam todos os parâmetros. Nos casos em que os traços biológicos medidos não apresentam fortes modificações devido à presença da enfermidade (ou infecção) no indivíduo, as estimativas podem chegar a tomar valores 50% maiores que o valor real, o que pode implicar importantes erros na tomada de decisões relacionadas à forma de tratar a doença / Abstract: Most studies to estimate the prevalence and performance clinical diagnostic test parameters have not considered that many of the diagnostic methods include the measurement of biological traits with outcome expressed in a continuous scale and that, due to these traits, are measured on the same individual, they necessarily have some kind of dependence that may or not be explained as a phenomenon of linear relation or agreement. Generally authors assume that the data have binary structure and they do not consider the fact the data take that form after of they be dichotomized using a cut-off. In this thesis, it is developed a proposal based on a Bayesian approach to the problem of estimating the prevalence, the sensitivity and the specificity of tests within designs that include the application of a gold standard for verification and two or three screening diagnostic tests each of them resulting from a measurement of a biological trait expressed in a continuous scale and dichotomization. Although the main objective of the proposed statistical model is to study the effect of dependence between test results on prevalence and performance test parameter estimates, it is also studied some alternatives found in the literature to address common difficulties in diagnostic test studies such as the lack of identifiability and the verification bias (specially when individuals with negative results in both screening tests are not verified by "gold standard"procedure). The proposed estimation method considers the use of three copula functions to study the effect of same number of dependence levels on the parameter test estimates. Since the study is based on a Bayesian approach, it was necessary to develop a procedure to eliciting prior distributions in situations of total absence of information about the parameter of interest, as is the case of copula functions that have been very little used in medical research. The obtained results are compared with those obtained using binary covariance and independence between test outcomes assumption, methods frequently used by other authors. Unlike the majority of methods presented in the literature on the matter, the proposed estimation model has a reparametrization that gives the estimates of interest directly without need of use complex analytical and computational methods and the results are easily obtained using a Winbugs 1.4. program. The dependence affects the estimates of the test parameters and it has little effect on the prevalence estimate. The effect is stronger in presence of verification bias. Weak dependences underestimate the sensitivities and other parameters are unbiased, while strong dependencies overestimate all parameters. In situations in which the biological traits measured did not show strong changes due to the presence of the disease (or infection) on the individual, the estimates can reach values close to 50% higher than the real value, which may involve important errors in decision making related to how to treat the disease / Doutorado / Estatistica e Probabilidade / Doutor em Estatística

Les travaux présentés s’inscrivent dans le cadre des études menées par la Caisse centrale de réassurance (CCR) pour modéliser les événements catastrophes naturelles et en particulier le péril sécheresse. La sécheresse est le nom utilisé couramment pour désigner les dommages aux bâtiments causés par le phénomène de retrait-gonflement des argiles. Les recherches effectuées sont en lien avec un modèle d’estimation du coût annuel d’un événement sécheresse conçu en interne à CCR. Celui-ci croise des données assurantielles et des données d’humidité du sol pour estimer le coût d’un événement survenu. CCR souhaite faire évoluer ce modèle vers une version probabiliste qui consiste à concevoir un générateur d’événements fictifs, non nécessairement survenus mais possibles d’un point de vue physique. Ce générateur doit permettre notamment d’estimer la distribution de probabilité du coût d’une sécheresse potentielle. Afin de concevoir un générateur d’événements fictifs pour le modèle sécheresse de CCR, nous avons étudié puis mis en oeuvre différents outils mathématiques permettant de modéliser la dépendance de variables aléatoires spatio-temporelles. La méthode choisie consiste à étudier et modéliser séparément la dépendance spatiale et la dépendance temporelle. Pour modéliser la dépendance temporelle, les modèles retenus sont des modèles classiques utilisés pour les séries temporelles. Nous décomposons les séries temporelles des observations en identifiant tendance et saisonnalité puis en ajustant un modèle autorégressif aux résidus. Pour modéliser la dépendance spatiale, notre choix s’est porté sur le krigeage et sur la théorie des copules. Les copules permettent de générer du bruit spatial pour ensuite lui appliquer les modèles de séries temporelles univariées. Le krigeage nous sert à interpoler spatialement les données générées dans le cas où une sélection de sites a été effectuée pour diminuer la dimension spatiale du problème. L’exploitation du générateur, pour laquelle nous donnons quelques résultats, va servir à CCR pour ses politiques de provisionnement et de tarification, et s’intègre également dans l’estimation de la charge deux-centennale liée aux catastrophes naturelles dans le cadre de la directive européenne Solvabilité II. / This work was performed at CCR, a French reinsurance company, within the studies that are conducted to model natural disasters, and particularly the drought hazard. Drought is the word used to denote the shrink-swell clay phenomenon that damages individual houses. These researches are related to an internal model that estimates the annual cost of a drought. This model crosses insurance data and soil moisture data to evaluate the cost of a occured event. CCR wants this model to be improved towards a probabilistic version by conceiving a generator of drought events that have to be realistic, although they are fictive. This generator will allow the estimation of the probability distribution of the drought cost. In order to conceive a fictive event generator for CCR’s drought model, mathematical tools have been used to model dependence between spatio-temporal random variables. The chosen method consists of studying and modeling separately spatial dependence and temporal dependence. Temporal dependence is modelized with time series models such as classical decomposition and autoregressive processes. Spatial dependence is modelized with kriging and copula theory. Spatial random noise is generated with a copula and then time series models are applied to rebuild original process. Kriging is used when generated data need to be interpolated, for example when data are generated only on a subset of the main grid. Results of the generator exploitation are given. They will be used by CCR for provisionning and pricing. These results will also be used for the estimation of the two-hundred-year cost of natural disasters within the new European Solvency II Directive.

Claims reserving and claims process estimation present classical problems in general insurance. The overall reserves are often determined under the assumption of independence among the lines of business. Though, recently modelling of the dependence among multiple lines of business has become crucial issue of reserving process. In this context, copulae provide a useful tool to construct models which go beyond the classical ones in terms of dependence structure. This thesis deals, in particular, with the copula regression model, its properties and possible applications in general insurance. This approach combines GLM modelling of margins and then expressing the dependence structure using copula. The theoretical methods are illustrated on a real dataset.

In non-life insurance, the independence between the number and size of claims is usually assumed. However, this thesis shows that the assumption of independence can be omitted. We deal with the dependency modeling between frequency and severity of claims. For including the dependence to the total claims model, we consider two methods. The first method uses generalized linear models and the second method used in the thesis is based on dependence modeling by copulas. We also perform a model with independent frequency and severity of claims. This model is compared with the described methods in the simulation part of the thesis. We include dependency on explanatory (rating) variables in all of these models. 1

Initialement, la modélisation des risques en assurance non vie, supposait l'indépendance entre les différentes variables des modèles actuariels. De nos jours, cette hypothèse d'indépendance est souvent relâchée afin de tenir compte de possibles interactions entre les différents éléments. Cette thèse a pour but de contribuer à la littérature existante de la modélisation de la dépendance en assurance non vie. Concrètement, nous introduisons une nouvelle méthodologie d'analyse des risques en assurance à travers le développement des modèles de dépendance, principalement dans un cadre dynamique. Dans le premier chapitre de la thèse nous introduisons le contexte actuel de solvabilité, ainsi que la modélisation de la dépendance en assurance, avec une présentation des principaux résultats. Le deuxième chapitre est essentiellement constitué d'un article coécrit avec Christian de Peretti et Lotfi Belkacem, intitulé "Modelling Temporal Dependence of Claims In Insurance Using Autoregressive Conditional Amount Models" (voir Araichi et al. (2013)). Dans ce chapitre nous montrons l'existence d'une forme de dépendance temporelle (dynamique) entre les montants de sinistres d'une même branche d'assurance. Nous proposons un nouveau modèle nommé Autoregressive Conditional Amount Model (ACA), qui permet de capturer le comportement dynamique des sinistres. Également, nous développons un nouveau modèle nommé Generalized Extreme Value ACA model (GEVACA), afin d'analyser la dépendance dynamique des montants élevés, au niveau des queues de distribution. Enfin, nous donnons une nouvelle expression pour la Value at Risk (VaR) paramétrique adaptée pour des risques à dépendance temporelle. Des applications sur des données réelles et des techniques de backtesting sont ensuite effectuées afin de montrer la pertinence des modèles proposés. Le troisième chapitre est constitué d'un article coécrit avec Christian de Peretti et Lotfi Belkacem, intitulé "Generalized Autoregressive Conditional Sinistrality Model : A novel model for claims reserving in Non life insurance", (voir Araichi et al. (2015)). Dans ce chapitre, nous abordons d'abord le problème de l'évaluation des réserves dans un cadre dynamique. Nous montrons l'existence d'une forme de dépendance dynamique dans un triangle de liquidation. En particulier, nous nous intéressons à l'analyse de la dépendance temporelle entre les sinistres, ainsi qu'entre les années de développement. Nous proposons un nouveau modèle nommé "Generalized Autoregressive Conditional Sinistrality Model (GACSM), qui constitue une extension du modèle linéaire généralisé classique. Ensuite, nous fournissons une méthode de simulation bootstrap basée sur le modèle GACSM, qui permet d'évaluer les réserves en tenant compte du caractère dynamique des sinistres. Enfin, afin de montrer l'impact du modèle proposé sur l'évaluation des réserves et du capital, nous effectuons une comparaison des résultats obtenus avec ceux obtenus des modèles classiques (Chain Ladder et modèle linéaire généralisé). Dans le quatrième chapitre de la thèse, qui est constitué d'un article, coécrit avec Christian de Peretti et Lotfi Belkacem, intitulé "Time Varying Copula Model for claims reserving in Non life insurance". Nous intéressons à évaluer le montant agrégé des sinistres, en analysant conjointement la dépendance dynamique inter-sinistres ainsi qu'entre les sinistres de deux branches. Nous proposons un modèle basé sur le modèle GACSM et les copules conditionnelles, qui permettent de suivre l'évolution de la dépendance au cours du temps. Enfin, nous effectuons des applications sur des données réelles, ainsi que des méthodes de simulation sont considérées. En comparant les résultats obtenus, nous avons pu illustrer l'impact de la dépendance dynamique sur les réserves et le besoin en capital / In this thesis a different aspects of dependence modeling are considered. Indeed, temporal dependence structures between claims amounts and between lines of business are analyzed. In the first chapter, a general introduction on modeling dependence in insurance is provided. The second chapter is essentially constituted by the article "Modeling Temporal Dependence of Claims In Insurance Using Autoregressive Conditional Amount Models", written with Christian de Peretti and Lotfi Belkacem, (see Araichi et al. (2013)) It deals with the problem of existing a temporal dependence structure between claims amounts of one line of business. To this end, we propose a new model for handling the dynamic behaviour of claims amounts in insurance companies using an Autoregressive Conditional Amount (ACA) framework. This model will be named Autoregressive Conditional Amount Model (ACA). A Gamma ACA model and a Generalized Extreme Value ACA model are proposed. It is shown that these models are more appropriate to describe and to forecast the process of claims of the lines Auto Damage and Auto Liability than traditional models. Furthermore, a parametric Value at Risk based on ACA framework (VaR ACA) is proposed for evaluating a coverage amount of these claims. Using backtesting techniques, the VaR ACA provides an accurate estimation of risk. The third chapter of this thesis is based on the article "Generalized Autoregressive Conditional Sinistrality Model: A novel model for claims reserving in Non life insurance", written with Christian de Peretti and Lotfi Belkacem, (see Araichi et al. (2015)). In this chapter, a Generalized Autoregressive Conditional Sinistrality Model (GACSM) for claims is proposed. We extend the Generalized Linear Model (GLM) by incorporating temporal dependence between claims amounts of one triangle. The GACSM is used for model parameter estimation and consistency of such estimate is proved. Bootstrap procedure is implemented for prediction reserves and prediction errors. Results show that taking into account the temporal dependence between losses improves the precision of the reserve distribution estimate, and thus evaluates an accurate SCR. Finally the fourth chapter is based on the article "Time Varying Copula Model for claims reserving in Non life insurance", written with Christian de Peretti and LotfiBelkacem. In this chapter, a time varying copula models to understand the behavior of claims amounts of two lines of business. Time varying copula functions with a Generalized Autoregressive Conditional Sinistrality model are used to analyze the evolution in time of dependence between two lines and the temporal dependence between claims of each line. Simulation study is performed to highlight the impact on reserves and Solvency Capital Requirement. Results show that our approach provides a diversification effect between claims amounts