Modeling Stock Order Flows and Learning Market-Making from Data

Stock markets employ specialized traders, market-makers, designed to provide liquidity and volume to the market by constantly supplying both supply and demand. In this paper, we demonstrate a novel method for modeling the market as a dynamic system and a reinforcement learning algorithm that learns profitable market-making strategies when run on this model. The sequence of buys and sells for a particular stock, the order flow, we model as an Input-Output Hidden Markov Model fit to historical data. When combined with the dynamics of the order book, this creates a highly non-linear and difficult dynamic system. Our reinforcement learning algorithm, based on likelihood ratios, is run on this partially-observable environment. We demonstrate learning results for two separate real stocks.

Valor en riesgo del portafolio de TES de los bancos colombianos

En este trabajo se realiza la medición del riesgo de mercado para el portafolio de TES de un banco colombiano determinado, abordando el pronóstico de valor en riesgo (VaR) mediante diferentes modelos multivariados de volatilidad: EWMA, GARCH ortogonal, GARCH robusto, así como distintos modelos de VaR con distribución normal y distribución t-student, evaluando su eficiencia con las metodologías de backtesting propuestas por Candelon et al. (2011) con base en el método generalizado de momentos, junto con los test de independencia y de cobertura condicional planteados por Christoffersen y Pelletier (2004) y por Berkowitz, Christoffersen y Pelletier (2010). Los resultados obtenidos demuestran que la mejor especificación del VaR para la medición del riesgo de mercado del portafolio de TES de los bancos colombianos, es el construido a partir de volatilidades EWMA y basado en la distribución normal, ya que satisface las hipótesis de cobertura no condicional, independencia y cobertura condicional, al igual que los requerimientos estipulados en Basilea II y en la normativa vigente en Colombia.

Media bias and central bank response. Evidence from the nominal exchange rate behavior in Colombia

Since 1991 Colombia has had a market-determined Peso - US Dollar Nominal Exchange Rate (NER), after more than 20 years of controlled and multiple exchange rates. The behavior (revaluation / devaluation) of the NER is constantly reported in news, editorials and op-eds of major newspapers of the nation with particular attention to revaluation. The uneven reporting of revaluation episodes can be explained by the existence of an interest group particulary affected by revaluation, looking to increase awareness and sympathy for help from public institutions. Using the number of news and op-eds from a major Colombian newspaper, it is shown that there is an over-reporting of revaluation episodes in contrast to devaluation ones. Secondly, using text analysis upon the content of the news, it is also shown that the words devaluation and revaluation are far apart in the distribution of words within the news; and revaluation is highly correlated with words related to: public institutions, exporters and the need of assistance. Finally it is also shown that the probability of the central bank buying US dollars to lessen revaluation effects increases with the number of news; even though the central bank allegedly intervenes in the exchange rate market only to tame volatility or accumulate international reserves.

Modeling the monetary policy reaction function of the colombian central bank

This paper proposes a simple Ordered Probit model to analyse the monetary policy reaction function of the Colombian Central Bank. There is evidence that the reaction function is asymmetric, in the sense that the Bank increases the Bank rate when the gap between observed inflation and the inflation target (lagged once) is positive, but it does not reduce the Bank rate when the gap is negative. This behaviour suggests that the Bank is more interested in fulfilling the announced inflation target rather than in reducing inflation excessively. The forecasting performance of the model, both within and beyond the estimation period, appears to be particularly good.

Dynamic conditional correlation in Latin-American asset markets

In this paper we reviewed the models of volatility for a group of five Latin American countries, mainly motivated by the recent periods of financial turbulence. Our results based on high frequency data suggest that Dynamic multivariate models are more powerful to study the volatilities of asset returns than Constant Conditional Correlation models. For the group of countries included, we identified that domestic volatilities of asset markets have been increasing; but the co-volatility of the region is still moderate.

Destabilizing Mutations Alter the Hydrogen Exchange Mechanism in Ribonuclease A

The effect of strongly destabilizing mutations, I106A and V108G of Ribonuclease A (RNase A), on its structure and stability has been determined by NMR. The solution structures of these variants are essentially equivalent to RNase A. The exchange rates of the most protected amide protons in RNase A (35ºC), the I106A variant (35ºC), and the V108G variant (10ºC) yield stability values of 9.9, 6.0, and 6.8 kcal/mol, respectively, when analyzed assuming an EX2 exchange mechanism. Thus, the destabilization induced by these mutations is propagated throughout the protein. Simulation of RNase A hydrogen exchange indicates that the most protected protons in RNase A and the V108G variant exchange via the EX2 regime, whereas those of I106A exchange through a mixed EX1 1 EX2 process. It is striking that a single point mutation can alter the overall exchange mechanism. Thus, destabilizing mutations joins high temperatures, high pH and the presence of denaturating agents as a factor that induces EX1 exchange in proteins. The calculations also indicate a shift from the EX2 to the EX1 mechanism for less protected groups within the same protein. This should be borne in mind when interpreting exchange data as a measure of local stability in less protected regions

A dynamic knowledge-based decision support system to handle solids separation problems in activated sludge systems: development and validation

El sistema de fangs activats és el tractament biològic més àmpliament utilitzat arreu del món per la depuració d'aigües residuals. El seu funcionament depèn de la correcta operació tant del reactor biològic com del decantador secundari. Quan la fase de sedimentació no es realitza correctament, la biomassa no decantada s'escapa amb l'efluent causant un impacte sobre el medi receptor. Els problemes de separació de sòlids, són actualment una de les principals causes d'ineficiència en l'operació dels sistemes de fangs activats arreu del món. Inclouen: bulking filamentós, bulking viscós, escumes biològiques, creixement dispers, flòcul pin-point i desnitrificació incontrolada. L'origen dels problemes de separació generalment es troba en un desequilibri entre les principals comunitats de microorganismes implicades en la sedimentació de la biomassa: els bacteris formadors de flòcul i els bacteris filamentosos. Degut a aquest origen microbiològic, la seva identificació i control no és una tasca fàcil pels caps de planta. Els Sistemes de Suport a la Presa de Decisions basats en el coneixement (KBDSS) són un grup d'eines informàtiques caracteritzades per la seva capacitat de representar coneixement heurístic i tractar grans quantitats de dades. L'objectiu de la present tesi és el desenvolupament i validació d'un KBDSS específicament dissenyat per donar suport als caps de planta en el control dels problemes de separació de sòlids d'orígen microbiològic en els sistemes de fangs activats. Per aconseguir aquest objectiu principal, el KBDSS ha de presentar les següents característiques: (1) la implementació del sistema ha de ser viable i realista per garantir el seu correcte funcionament; (2) el raonament del sistema ha de ser dinàmic i evolutiu per adaptar-se a les necessitats del domini al qual es vol aplicar i (3) el raonament del sistema ha de ser intel·ligent. En primer lloc, a fi de garantir la viabilitat del sistema, s'ha realitzat un estudi a petita escala (Catalunya) que ha permès determinar tant les variables més utilitzades per a la diagnosi i monitorització dels problemes i els mètodes de control més viables, com la detecció de les principals limitacions que el sistema hauria de resoldre. Els resultats d'anteriors aplicacions han demostrat que la principal limitació en el desenvolupament de KBDSSs és l'estructura de la base de coneixement (KB), on es representa tot el coneixement adquirit sobre el domini, juntament amb els processos de raonament a seguir. En el nostre cas, tenint en compte la dinàmica del domini, aquestes limitacions es podrien veure incrementades si aquest disseny no fos òptim. En aquest sentit, s'ha proposat el Domino Model com a eina per dissenyar conceptualment el sistema. Finalment, segons el darrer objectiu referent al seguiment d'un raonament intel·ligent, l'ús d'un Sistema Expert (basat en coneixement expert) i l'ús d'un Sistema de Raonament Basat en Casos (basat en l'experiència) han estat integrats com els principals sistemes intel·ligents encarregats de dur a terme el raonament del KBDSS. Als capítols 5 i 6 respectivament, es presenten el desenvolupament del Sistema Expert dinàmic (ES) i del Sistema de Raonament Basat en Casos temporal, anomenat Sistema de Raonament Basat en Episodis (EBRS). A continuació, al capítol 7, es presenten detalls de la implementació del sistema global (KBDSS) en l'entorn G2. Seguidament, al capítol 8, es mostren els resultats obtinguts durant els 11 mesos de validació del sistema, on aspectes com la precisió, capacitat i utilitat del sistema han estat validats tant experimentalment (prèviament a la implementació) com a partir de la seva implementació real a l'EDAR de Girona. Finalment, al capítol 9 s'enumeren les principals conclusions derivades de la present tesi.

Front spreading in population dynamic models. Theory and application to the Neolithic transition

This thesis presents population dynamics models that can be applied to predict the rate of spread of the Neolithic transition (change from hunter-gathering to farming economics) across the European continent, which took place about 9000 to 5000 years ago. The first models in this thesis provide predictions at a continental scale. We develop population dynamics models with explicit kernels and apply realistic data. We also derive a new time-delayed reaction-diffusion equation which yields speeds about a 10% slower than previous models. We also deal with a regional variability: the slowdown of the Neolithic front when reaching the North of Europe. We develop simple reaction-diffusion models that can predict the measured speeds in terms of the non-homogeneous distribution of pre-Neolithic (Mesolithic) population in Europe, which were present in higher densities at the North of the continent. Such models can explain the observed speeds.

Integration of knowledge-based, qualitative and numeric tools for real time dynamic systems supervision

The proposal presented in this thesis is to provide designers of knowledge based supervisory systems of dynamic systems with a framework to facilitate their tasks avoiding interface problems among tools, data flow and management. The approach is thought to be useful to both control and process engineers in assisting their tasks. The use of AI technologies to diagnose and perform control loops and, of course, assist process supervisory tasks such as fault detection and diagnose, are in the scope of this work. Special effort has been put in integration of tools for assisting expert supervisory systems design. With this aim the experience of Computer Aided Control Systems Design (CACSD) frameworks have been analysed and used to design a Computer Aided Supervisory Systems (CASSD) framework. In this sense, some basic facilities are required to be available in this proposed framework: ·

A general-method for tracking analysis and its application to meteorological data

Recent interest in the validation of general circulation models (GCMs) has been devoted to objective methods. A small number of authors have used the direct synoptic identification of phenomena together with a statistical analysis to perform the objective comparison between various datasets. This paper describes a general method for performing the synoptic identification of phenomena that can be used for an objective analysis of atmospheric, or oceanographic, datasets obtained from numerical models and remote sensing. Methods usually associated with image processing have been used to segment the scene and to identify suitable feature points to represent the phenomena of interest. This is performed for each time level. A technique from dynamic scene analysis is then used to link the feature points to form trajectories. The method is fully automatic and should be applicable to a wide range of geophysical fields. An example will be shown of results obtained from this method using data obtained from a run of the Universities Global Atmospheric Modelling Project GCM.

Calculating the ocean's mean dynamic topography from a mean sea surface and a geoid

In principle the global mean geostrophic surface circulation of the ocean can be diagnosed by subtracting a geoid from a mean sea surface (MSS). However, because the resulting mean dynamic topography (MDT) is approximately two orders of magnitude smaller than either of the constituent surfaces, and because the geoid is most naturally expressed as a spectral model while the MSS is a gridded product, in practice complications arise. Two algorithms for combining MSS and satellite-derived geoid data to determine the ocean’s mean dynamic topography (MDT) are considered in this paper: a pointwise approach, whereby the gridded geoid height field is subtracted from the gridded MSS; and a spectral approach, whereby the spherical harmonic coefficients of the geoid are subtracted from an equivalent set of coefficients representing the MSS, from which the gridded MDT is then obtained. The essential difference is that with the latter approach the MSS is truncated, a form of filtering, just as with the geoid. This ensures that errors of omission resulting from the truncation of the geoid, which are small in comparison to the geoid but large in comparison to the MDT, are matched, and therefore negated, by similar errors of omission in the MSS. The MDTs produced by both methods require additional filtering. However, the spectral MDT requires less filtering to remove noise, and therefore it retains more oceanographic information than its pointwise equivalent. The spectral method also results in a more realistic MDT at coastlines. 1. Introduction An important challenge in oceanography is the accurate determination of the ocean’s time-mean dynamic topography (MDT). If this can be achieved with sufficient accuracy for combination with the timedependent component of the dynamic topography, obtainable from altimetric data, then the resulting sum (i.e., the absolute dynamic topography) will give an accurate picture of surface geostrophic currents and ocean transports.

Language processing with dynamic fields

We construct a mapping from complex recursive linguistic data structures to spherical wave functions using Smolensky's filler/role bindings and tensor product representations. Syntactic language processing is then described by the transient evolution of these spherical patterns whose amplitudes are governed by nonlinear order parameter equations. Implications of the model in terms of brain wave dynamics are indicated.