Résumé : Cette thèse comporte 8 chapitres. Le chapitre 1 est une introduction aux problématiques rencontrées sur les marchés énergétiques : fréquence d'intervention faible, coûts de transaction élevés, évaluation des options spread. Le chapitre 2 étudie la convergence de l'erreur de couverture d'une option call dans le modèle de Bachelier, pour des coûts de transaction proportionnels (modèle de Leland-Lott) et lorsque la fréquence d'intervention devient infinie. Il est prouvé que cette erreur est bornée par une variable aléatoire proportionnelle au taux de transaction. Cependant, les démonstrations de convergence en probabilité demandent des régularités sur les sensibilités assez restrictives en pratique. Les chapitres suivants contournent ces obstacles en étudiant des convergences presque sûres. Le chapitre 3 développe tout d'abord de nouveaux outils de convergence presque sûre. Ces résultats ont de nombreuses conséquences sur le contrôle presque sûr de martingales et de leur variation quadratique, ainsi que de leurs incréments entre deux temps d'arrêt généraux. Ces résultats de convergence trajectorielle sont connus pour être difficiles à obtenir sans information sur les lois. Par la suite, nous appliquons ces résultats à la minimisation presque sûre de la variation quadratique renormalisée de l'erreur de couverture d'une option de payoff général (cadre multidimensionnel, payoff asiatique, lookback) sur une large classe de temps d'intervention. Une borne inférieure à notre critère est trouvée et une suite minimisante de temps d'arrêt optimale est exhibée : il s'agit de temps d'atteinte d'ellipsoïde aléatoire, dépendant du gamma de l'option. Le chapitre 4 étudie la convergence de l'erreur de couverture d'une option de payoff convexe (dimension 1) en prenant en compte des coûts de transaction à la Leland-Lott. Nous décomposons l'erreur de couverture en une partie martingale et une partie négligeable, puis nous minimisons la variation quadratique de cette martingale sur une classe de temps d'atteintes générales pour des Deltas vérifiant une certaine EDP non-linéaire sur les dérivées secondes. Nous exhibons aussi une suite de temps d'arrêt atteignant cette borne. Des tests numériques illustrent notre approche par rapport à une série de stratégies connues de la littérature. Le chapitre 5 étend le chapitre 3 en considérant une fonctionnelle des variations discrètes d'ordre Y et de Z de deux processus d'Itô Y et Z à valeurs réelles, la minimisation étant sur une large classe de temps d'arrêt servant au calcul des variations discrètes. Borne inférieure et suite minimisant sont obtenues. Une étude numérique sur les coûts de transaction est faite. Le chapitre 6 étudie la discrétisation d'Euler d'un processus multidimensionnel X dirigé par une semi-martingale d'Itô Y . Nous minimisons sur les temps de la grille de discrétisation un critère quadratique sur l'erreur du schéma. Nous trouvons une borne inférieure et une grille optimale, ne dépendant que des données observables. Le chapitre 7 donne un théorème limite centrale pour des discrétisations d'intégrale stochastique sur des grilles de temps d'atteinte d'ellipsoïdes adaptées quelconque. La corrélation limite est conséquence d'asymptotiques fins sur les problèmes de Dirichlet. Dans le chapitre 8, nous nous intéressons aux formules d'expansion pour les options sur spread, pour des modèles à volatilité locale. La clé de l'approche consiste à conserver la propriété de martingale de la moyenne arithmétique et à exploiter la structure du payoff call. Les tests numériques montrent la pertinence de l'approche.

Nesta dissertação de mestrado nós trabalhamos com o processo de Cox-Ingersoll- Ross, que foi originalmente proposto por John C. Cox, Jonathan E. Ingersoll Jr. e Stephen A. Ross em 1985. Este processo é amplamente utilizado em modelagem financeira, por exemplo, para descrever a evolução de taxas de juros ou como o processo de volatilidade no modelo de Heston. A equação diferencial estocástica que define este processo não possui solução fechada, logo faz-se necessária a aproximação do processo via algum método numérico. Na literatura diversos trabalhos propõem aproximações baseadas em esquemas de discretização intervalar. Nós aproximamos o processo de Cox-Ingersoll-Ross através de um método numérico do tipo Euler- Maruyama baseado na discretização aleatória proposta por Leão e Ohashi (2013) sob a condição de Feller. Neste contexto, mostramos que esta aproximação possui uma ordem de convergência exponencial e utilizamos técnicas de simulação Monte Carlo para comparar resultados numéricos com valores teóricos. / In this master\'s thesis we work with Cox-Ingersoll-Ross (CIR) process. This process was originally proposed by John C. Cox, Jonathan E. Ingersoll Jr. and Stephen A. Ross in 1985. Nowadays, this process is widely used in financial modeling, e.g. as a model for short-time interest rates or as volatility process in the Heston model. The stochastic differential equation (SDE) which defines this model does not have closed form solution, so we need to approximate the process by some numerical method. In the literature, several numerical approximations has been proposed based in interval discretization. We approximate the CIR process by Euler-Maruyama-type method based in random discretization proposed by Leão e Ohashi (2013) under Feller condition. In this context, we obtain an exponential convergence order for this approximation and we use Monte Carlo techniques to compare the numerical results with theoretical values.

A standard problem in mathematical finance is the calculation of the price of some financial derivative such as various types of options. Since there exists analytical solutions in only a few cases it will often boil down to estimating the price with Monte Carlo simulation in conjunction with some numerical discretization scheme. The upside of using what we can call standard Monte Carlo is that it is relative straightforward to apply and can be used for a wide variety of problems. The downside is that it has a relatively slow convergence which means that the computational cost or complexity can be very large. However, this slow convergence can be improved upon by using Multilevel Monte Carlo instead of standard Monte Carlo. With this approach it is possible to reduce the computational complexity and cost of simulation considerably. The aim of this thesis is to introduce the reader to the Multilevel Monte Carlo method with applications to European and Asian call options in both the Black-Scholes-Merton (BSM) model and in the Heston model. To this end we first cover the necessary background material such as basic probability theory, estimators and some of their properties, the stochastic integral, stochastic processes and Ito’s theorem. We introduce stochastic differential equations and two numerical discretizations schemes, the Euler–Maruyama scheme and the Milstein scheme. We define strong and weak convergence and illustrate these concepts with examples. We also describe the standard Monte Carlo method and then the theory and implementation of Multilevel Monte Carlo. In the applications part we perform numerical experiments where we compare standard Monte Carlo to Multilevel Monte Carlo in conjunction with the Euler–Maruyama scheme and Milsteins scheme. In the case of a European call in the BSM model, using the Euler–Maruyama scheme, we achieved a cost O(ε-2(log ε)2) to reach the desired error in accordance with theory in comparison to the O(ε-3) cost for standard Monte Carlo. When using Milsteins scheme instead of the Euler–Maruyama scheme it was possible to reduce the cost in terms of the number of simulations needed to achieve the desired error even further. By using Milsteins scheme, a method with greater order of strong convergence than Euler–Maruyama, we achieved the O(ε-2) cost predicted by the complexity theorem compared to the standard Monte Carlo cost of order O(ε-3). In the final numerical experiment we applied the Multilevel Monte Carlo method together with the Euler–Maruyama scheme to an Asian call in the Heston model. In this case, where the coefficients of the Heston model do not satisfy a global Lipschitz condition, the study of strong or weak convergence is much harder. The numerical experiments suggested that the strong convergence was slightly slower compared to what was found in the case of a European call in the BSM model. Nevertheless, we still achieved substantial savings in computational cost compared to using standard Monte Carlo.

Many phenomena in our world can be described as differential equations in high dimensions. However, they are notoriously challenging to solve numerically due to the exponential growth in computational cost with increasing dimensions. This thesis explores an algorithm, known as deep BSDE, for solving high-dimensional partial differential equations and applies it to finance, namely European option pricing. In addition, an implementation of the method is provided that seemingly shortens the runtime by a factor of two, compared with the results in previous studies. From the results, we can conclude that the deep BSDE method does handle high-dimensional problems well. Lastly, the thesis gives the relevant prerequisites required to be able to digest the theory from an undergraduate level.

This thesis investigated Stochastic Runge–Kutta Lawson (SRKL) schemes and their application to the Heston model. Two distinct SRKL discretization methods were used to simulate a single asset’s dynamics under the Heston model, notably the Euler–Maruyama and Midpoint schemes. Additionally, standard Monte Carlo and variance reduction techniques were implemented. European and Asian option prices were estimated and compared with a benchmark value regarding accuracy, effectiveness, and computational complexity. Findings showed that the SRKL Euler–Maruyama schemes exhibited promise in enhancing the price for simple and path-dependent options. Consequently, integrating SRKL numerical methods into option valuation provides notable advantages by addressing challenges posed by the Heston model’s SDEs. Given the limited scope of this research topic, it is imperative to conduct further studies to understand the use of SRKL schemes within other models.