During the last few years, market micro-structure research has been active in analysing the dependence of market efficiency on different market character­ istics. Make-take fees are one of those topics as they might modify the incen­ tives for participating agents, e.g. broker-dealers or market-makers. In this thesis, we propose a Hawkes process-based model that captures statistical differences arising from different fee regimes and we estimate the differences on limit order book data. We then use these estimates in an attempt to measure the execution quality from the perspective of a market-maker. We appropriate existing theoretical market frameworks, however, for the pur­ pose of hireling optimal market-making policies we apply a novel method of deep reinforcement learning. Our results suggest, firstly, that maker-taker exchanges provide better liquidity to the markets, and secondly, that deep reinforcement learning methods may be successfully applied to the domain of optimal market-making. JEL Classification Keywords Author's e-mail Supervisor's e-mail C32, C45, C61, C63 make-take fees, Hawkes process, limit order book, market-making, deep reinforcement learn­ ing kiselrastislavSgmail.com barunik@f sv.cuni.cz

<p>As more and more datasets with self-exciting properties become available, the demand for robust models that capture contagion across events is also getting stronger. Hawkes processes stand out given their ability to capture a wide range of contagion and self-excitation patterns, including the transmission of infectious disease, earthquake aftershock distributions, near-repeat crime patterns, and overdose clusters. The Hawkes process is flexible in modeling these various applications through parametric and non-parametric kernels that model event dependencies in space, time and on networks.</p> <p>In this thesis, we develop new frameworks that integrate Hawkes Process models with multi-armed bandit algorithms, high dimensional marks, and high-dimensional auxiliary data to solve problems in search and rescue, forecasting infectious disease, and early detection of overdose spikes.</p> <p>In Chapter 3, we develop a method applications to the crisis of increasing overdose mortality over the last decade.  We first encode the molecular substructures found in a drug overdose toxicology report. We then cluster these overdose encodings into different overdose categories and model these categories with spatio-temporal multivariate Hawkes processes. Our results demonstrate that the proposed methodology can improve estimation of the magnitude of an overdose spike based on the substances found in an initial overdose. </p> <p>In Chapter 4, we build a framework for multi-armed bandit problems arising in event detection where the underlying process is self-exciting. We derive the expected number of events for Hawkes processes given a parametric model for the intensity and then analyze the regret bound of a Hawkes process UCB-normal algorithm. By introducing the Hawkes Processes modeling into the upper confidence bound construction, our models can detect more events of interest under the multi-armed bandit problem setting. We apply the Hawkes bandit model to spatio-temporal data on crime events and earthquake aftershocks. We show that the model can quickly learn to detect hotspot regions, when events are unobserved, while striking a balance between exploitation and exploration. </p> <p>In Chapter 5, we present a new spatio-temporal framework for integrating Hawkes processes with multi-armed bandit algorithms. Compared to the methods proposed in Chapter 4, the upper confidence bound is constructed through Bayesian estimation of a spatial Hawkes process to balance the trade-off between exploiting and exploring geographic regions. The model is validated through simulated datasets and real-world datasets such as flooding events and improvised explosive devices (IEDs) attack records. The experimental results show that our model outperforms baseline spatial MAB algorithms through rewards and ranking metrics.</p> <p>In Chapter 6, we demonstrate that the Hawkes process is a powerful tool to model the infectious disease transmission. We develop models using Hawkes processes with spatial-temporal covariates to forecast COVID-19 transmission at the county level. In the proposed framework, we show how to estimate the dynamic reproduction number of the virus within an EM algorithm through a regression on Google mobility indices. We also include demographic covariates as spatial information to enhance the accuracy. Such an approach is tested on both short-term and long-term forecasting tasks. The results show that the Hawkes process outperforms several benchmark models published in a public forecast repository. The model also provides insights on important covariates and mobility that impact COVID-19 transmission in the U.S.</p> <p>Finally, in chapter 7, we discuss implications of the research and future research directions.</p>

We aim at generalizing the celebrated portfolio optimization problem "à la Merton", where the asset evolution is steered by a self-exciting jump-diffusion process. We first define the rigorous mathematical framework needed to introduce the stochastic optimal control problem we are interesting in. Then, we provide a proof for a specific version of the Dynamic Programming Principle (DPP) with respect to the general class of self-exciting processes under study. After, we state the Hamilton-Jacobi-Bellman (HJB) equation, whose solution gives the value function for the corresponding optimal control problem. The resulting HJB equation takes the form of a Partial-Integro Differential Equation (PIDE), for which we prove both existence and uniqueness for the solution in the viscosity sense. We further derive a suitable numerical scheme to solve the HJB equation corresponding to the portfolio optimizationproblem. To this end, we also provide a detailed study of solution dependence on the parameters of the problem. The analysis is performed by calibrating the model on ENI asset levels during the COVID-19 worldwide breakout. In particular, the calibration routine is based on a sophisticated Sequential Monte Carlo algorithm.

L'objet de cette thèse est d'étudier divers problèmes de statistique non-paramétrique dans le cadre d'un modèle d'interactions poissoniennes. De tels modèles sont, par exemple, utilisés en neurosciences pour analyser les interactions entre deux neurones au travers leur émission de potentiels d'action au cours de l'enregistrement de l'activité cérébrale ou encore en génomique pour étudier les distances favorisées ou évitées entre deux motifs le long du génome. Dans ce cadre, nous introduisons une fonction dite de reproduction qui permet de quantifier les positions préférentielles des motifs et qui peut être modélisée par l'intensité d'un processus de Poisson. Dans un premier temps, nous nous intéressons à l'estimation de cette fonction que l'on suppose très localisée. Nous proposons une procédure d'estimation adaptative par seuillage de coefficients d'ondelettes qui est optimale des points de vue oracle et minimax. Des simulations et une application en génomique sur des données réelles provenant de la bactérie E. coli nous permettent de montrer le bon comportement pratique de notre procédure. Puis, nous traitons les problèmes de test associés qui consistent à tester la nullité de la fonction de reproduction. Pour cela, nous construisons une procédure de test optimale du point de vue minimax sur des espaces de Besov faibles, qui a également montré ses performances du point de vue pratique. Enfin, nous prolongeons ces travaux par l'étude d'une version discrète en grande dimension du modèle précédent en proposant une procédure adaptative de type Lasso. / The subject of this thesis is the study of some adaptive nonparametric statistical problems in the framework of a Poisson interactions model. Such models are used, for instance, in neurosciences to analyze interactions between two neurons through their spikes emission during the recording of the brain activity or in genomics to study favored or avoided distances between two motifs along a genome. In this setting, we naturally introduce a so-called reproduction function that allows to quantify the favored positions of the motifs and which is considered as the intensity of a Poisson process. Our first interest is the estimation of this function assumed to be well localized. We propose a data-driven wavelet thresholding estimation procedure that is optimal from oracle and minimax points of view. Simulations and an application to genomic data from the bacterium E. coli allow us to show the good practical behavior of our procedure. Then, we deal with associated problems on tests which consist in testing the nullity of the reproduction function. For this purpose, we build a minimax optimal testing procedure on weak Besov spaces and we provide some simulations showing good practical performances of our procedure. Finally, we extend this work with the study of a high-dimensional discrete setting of our previous model by proposing an adaptive Lasso-type procedure.

L'objet de cette thèse est d'étudier divers problèmes de statistique non-paramétrique dans le cadre d'un modèle d'interactions poissoniennes. De tels modèles sont, par exemple, utilisés en neurosciences pour analyser les interactions entre deux neurones au travers leur émission de potentiels d'action au cours de l'enregistrement de l'activité cérébrale ou encore en génomique pour étudier les distances favorisées ou évitées entre deux motifs le long du génome. Dans ce cadre, nous introduisons une fonction dite de reproduction qui permet de quantifier les positions préférentielles des motifs et qui peut être modélisée par l'intensité d'un processus de Poisson. Dans un premier temps, nous nous intéressons à l'estimation de cette fonction que l'on suppose très localisée. Nous proposons une procédure d'estimation adaptative par seuillage de coefficients d'ondelettes qui est optimale des points de vue oracle et minimax. Des simulations et une application en génomique sur des données réelles provenant de la bactérie E. coli nous permettent de montrer le bon comportement pratique de notre procédure. Puis, nous traitons les problèmes de test associés qui consistent à tester la nullité de la fonction de reproduction. Pour cela, nous construisons une procédure de test optimale du point de vue minimax sur des espaces de Besov faibles, qui a également montré ses performances du point de vue pratique. Enfin, nous prolongeons ces travaux par l'étude d'une version discrète en grande dimension du modèle précédent en proposant une procédure adaptative de type Lasso.

Les processus ponctuels sont souvent utilisés comme modèles de répartitions spatiales ou spatio-temporelles d'occurrences. Dans cette thèse, nous nous intéressons tout d'abord à des processus de Cox dirigés par un processus caché associé à un processus de Dirichlet. Ce modèle correspond à des occurrences cachées influençant l'intensité stochastique des occurrences observées. Nous généralisons la notion de " Shot noise Cox process " introduite par Moller et développons le traitement bayésien par un échantillonneur de Gibbs combiné à un algorithme de Metropolis-Hastings. Nous montrons que cette méthode MCMC est à sauts réversibles. Le modèle prend en compte, en effet, un nombre aléatoire de contributions cachées influençant l'intensité du processus ponctuel observé donc a un espace paramétrique de dimension variable. Nous focalisons l'inférence statistique sur l'estimation de la valeur espérée de chaque contribution cachée, le nombre espéré de contributions cachées, le degré d'influence spatiale de ces contributions et leur degré de corrélation. Le test d'égalité des contributions et celui de leur indépendance sont ainsi développés. L'utilité en épidémiologie et en écologie est alors démontrée à partir de données de Rubus fruticosa, Ibicella lutea et de mortalité dans les cantons de Georgia, USA. En termes de données observées, deux situations sont considérées: premièrement, les positions spatiales des occurrences sont observées entre plusieurs paires de dates consécutives; deuxièmement, des comptages sont effectués, au cours d'une période fixée, dans des unités d'échantillonnage spatiales. D'autre part, nous nous intéressons aux processus ponctuels à mémoire introduits par Kagan, Ogata et Vere-Jones, précurseurs de la statistique sismologique. En effet, les processus ponctuels spatio-temporels ont une place importante dans l'étude des catalogues sismiques puisque ces derniers sont généralement constitués d'événements sismiques datés et géo-référencés. Nous avons étudié un modèle ETAS (Epidemic Type Aftershock Sequence) avec une intensité d'arrière-plan indépendante du temps et plusieurs fonctions déclenchantes permettant d'intégrer les événements antérieurs récents. Cette approche est utilisée pour étudier la sismicité de l'arc des Petites Antilles. Une étude comparative des modèles Gamma, Weibull, Log-Normal et loi d'Omori modifiée pour les fonctions déclenchantes est menée. Nous montrons que la loi d'Omori modifiée ne s'ajuste pas aux données sismiques des Petites Antilles et la fonction déclenchante la plus adaptée est le modèle de Weibull. Cela implique que le temps d'attente entre répliques dans la zone des Petites Antilles est plus faible que celui des régions à sismicité décrite par la loi d'Omori modifiée. Autrement dit, l'agrégation des répliques après un événement majeur est plus prononcée dans la zone des Petites Antilles. La possibilité d'inclure une intensité d'arrière-plan suivant un processus de Dirichlet centré sur un processus spatial log-gaussien est discutée.

<div><div><div><p>Many complex processes can be viewed as sequential events on a network. In this thesis, we study the interplay between a network and the event sequences on it. We first focus on predicting events on a known network. Examples of such include: modeling retweet cascades, forecasting earthquakes, and tracing the source of a pandemic. In specific, given the network structure, we solve two types of problems - (1) forecasting future events based on the historical events, and (2) identifying the initial event(s) based on some later observations of the dynamics. The inverse problem of inferring the unknown network topology or links, based on the events, is also of great important. Examples along this line include: constructing influence networks among Twitter users from their tweets, soliciting new members to join an event based on their participation history, and recommending positions for job seekers according to their work experience. Following this direction, we study two types of problems - (1) recovering influence networks, and (2) predicting links between a node and a group of nodes, from event sequences.</p></div></div></div>

No description available.

An optimal dividend problem with investment opportunities, taking into consideration a source of strategic risk is being considered, as well as the effect of market frictions on the decision process of the financial entities. It concerns the problem of determining an optimal control of the dividend under debt constraints and investment opportunities in an economy with business cycles. It is assumed that the company is to be allowed to accept or reject investment opportunities arriving at random times with random sizes, by changing its outstanding indebtedness, which would impact its capital structure and risk profile. This work mainly focuses on the strategic risk faced by the companies; and, in particular, it focuses on the manager's problem of setting appropriate priorities to deploy the limited resources available. This component is taken into account by introducing frictions in the capital structure modification process. The problem is formulated as a bi-dimensional singular control problem under regime switching in presence of jumps. An explicit condition is obtained in order to ensure that the value function is finite. A viscosity solution approach is used to get qualitative descriptions of the solution. Moreover, a lending scheme for a system of interconnected banks with probabilistic constraints of failure is being considered. The problem arises from the fact that financial institutions cannot possibly carry enough capital to withstand counterparty failures or systemic risk. In such situations, the central bank or the government becomes effectively the risk manager of last resort or, in extreme cases, the lender of last resort. If, on the one hand, the health of the whole financial system depends on government intervention, on the other hand, guaranteeing a high probability of salvage may result in increasing the moral hazard of the banks in the financial network. A closed form solution for an optimal control problem related to interbank lending schemes has been derived, subject to terminal probability constraints on the failure of banks which are interconnected through a financial network. The derived solution applies to real bank networks by obtaining a general solution when the aforementioned probability constraints are assumed for all the banks. We also present a direct method to compute the systemic relevance parameter for each bank within the network. Finally, a possible computation technique for the Default Risk Charge under to regulatory risk measurement processes is being considered. We focus on the Default Risk Charge measure as an effective alternative to the Incremental Risk Charge one, proposing its implementation by a quasi exhaustive-heuristic algorithm to determine the minimum capital requested to a bank facing the market risk associated to portfolios based on assets emitted by several financial agents. While most of the banks use the Monte Carlo simulation approach and the empirical quantile to estimate this risk measure, we provide new computational approaches, exhaustive or heuristic, currently becoming feasible, because of both new regulation and the high speed - low cost technology available nowadays.