In this thesis we present a numerical and analytical study of modelling extremes in chaotic dynamical systems. We study a range of examples with different dependency structures, and different clustering characteristics. We compare our analysis to the extreme statistics observed for financial returns data, and hence consider the modelling potential of using chaotic systems for understanding financial returns. As part of the study we use the block maxima approach and the peak over threshold method to compute the distribution parameters that arise in the corresponding extreme value distributions. We compare these computations to the theoretical answers, and moreover we obtain error bounds on the rate of convergence of these schemes. In particular we investigate the optimal block size when applying the block maxima method. Since the time series of observations on a dynamical system have dependency we must therefore go beyond the classic approach of studying extremes for independent identically distributed random variables. This is the main purpose of our study. As part of this thesis, we also study clustering in financial returns, and again investigate the potential of using dynamical systems models. Moreover we can also compare numerical quantification of clustering with theoretical approaches. As further work, we measure the dependency structures in our models using a rescaled range analysis. We also make preliminary investigations into record statistics for dynamical systems models, and relate our findings to record statistics in financial data, and to other models (such as random walk models).

Orientador: José Roberto Campanha / Banca: Anderson Luis Hebling Christofoletti / Banca: Osvaldo Missiato / Os sinais analisados são séries temporais de precipitações pluviométricas ou simplesmente denominadas chuvas, que sofrem influências de outras variáveis atmosféricas, como a temperatura, pressão, vento, relevo, posição geográfica, sazonalidade, dentre outras, constituindo um sistema complexo. Estas séries temporais de chuvas, foram obtidas de 48 postos de coleta de dados, com medidas diária, em (mm), de quantidade de chuva, pertencentes a 38 municípios, localizados nas 9 regiões climáticas do Estado de São Paulo, proposto por Monteiro (1973). Os valores do expoente de Hurst, destas séries temporais, foram estimados com o método conhecido como análise R/S, o método utilizando a transformada de Fourier e o método utilizando a transformada de wavelets. A análise R/S e o método utilizando a transformada de Fourier apresentaram resultados equivalentes, mostrando coerência e grande importância na análise de sistemas complexos, objeto deste estudo. O método utilizando a transformada de wavelets, forneceu alguns resultados coerentes, uma grande parte, com resultados superestimados e uma pequena parte, com resultados subestimados, em relação aos outros dois métodos, mostrando-se inadequado para esta análise. / We analyze temporal series associated to pluvial precipitations, best known as rain, The latter depends on temperature, pressure, landscape, location and season, among many other atmospheric variables, thus qualifying as a complex system. These rain's temporal series were obtained from 48 data collection posts, with daily rain measurements (in mm), associated to 38 counties within the nine climatic regions of the State of São Paulo. The values of the time series Hurst exponent, were computed by three methods namely: the R/S analysis method, a Fourier transform and the wavelet transform method. The first two yield coherent results, showing both the consistency and relevance of these methods when applied to complex systems, the main goal of this work. The wavelet method yielded higher and lower values for the Hurst exponent, thus probing the limitations of this method. / Mestre

Estimation of the self-similarity exponent has attracted growing interest in recent decades and became a research subject in various fields and disciplines. Real-world data exhibiting self-similar behavior and/or parametrized by self-similarity exponent (in particular Hurst exponent) have been collected in different fields ranging from finance and human sciencies to hydrologic and traffic networks. Such rich classes of possible applications obligates researchers to investigate qualitatively new methods for estimation of the self-similarity exponent as well as identification of long-range dependencies (or long memory). In this thesis I present the Bayesian estimation of the Hurst exponent. In contrast to previous methods, the Bayesian approach allows the possibility to calculate the point estimator and confidence intervals at the same time, bringing significant advantages in data-analysis as discussed in this thesis. Moreover, it is also applicable to short data and unevenly sampled data, thus broadening the range of systems where the estimation of the Hurst exponent is possible. Taking into account that one of the substantial classes of great interest in modeling is the class of Gaussian self-similar processes, this thesis considers the realizations of the processes of fractional Brownian motion and fractional Gaussian noise. Additionally, applications to real-world data, such as the data of water level of the Nile River and fixational eye movements are also discussed. / Die Abschätzung des Selbstähnlichkeitsexponenten hat in den letzten Jahr-zehnten an Aufmerksamkeit gewonnen und ist in vielen wissenschaftlichen Gebieten und Disziplinen zu einem intensiven Forschungsthema geworden. Reelle Daten, die selbsähnliches Verhalten zeigen und/oder durch den Selbstähnlichkeitsexponenten (insbesondere durch den Hurst-Exponenten) parametrisiert werden, wurden in verschiedenen Gebieten gesammelt, die von Finanzwissenschaften über Humanwissenschaften bis zu Netzwerken in der Hydrologie und dem Verkehr reichen. Diese reiche Anzahl an möglichen Anwendungen verlangt von Forschern, neue Methoden zu entwickeln, um den Selbstähnlichkeitsexponenten abzuschätzen, sowie großskalige Abhängigkeiten zu erkennen. In dieser Arbeit stelle ich die Bayessche Schätzung des Hurst-Exponenten vor. Im Unterschied zu früheren Methoden, erlaubt die Bayessche Herangehensweise die Berechnung von Punktschätzungen zusammen mit Konfidenzintervallen, was von bedeutendem Vorteil in der Datenanalyse ist, wie in der Arbeit diskutiert wird. Zudem ist diese Methode anwendbar auf kurze und unregelmäßig verteilte Datensätze, wodurch die Auswahl der möglichen Anwendung, wo der Hurst-Exponent geschätzt werden soll, stark erweitert wird. Unter Berücksichtigung der Tatsache, dass der Gauß'sche selbstähnliche Prozess von bedeutender Interesse in der Modellierung ist, werden in dieser Arbeit Realisierungen der Prozesse der fraktionalen Brown'schen Bewegung und des fraktionalen Gauß'schen Rauschens untersucht. Zusätzlich werden Anwendungen auf reelle Daten, wie Wasserstände des Nil und fixierte Augenbewegungen, diskutiert.

Turbochargers are used on many automotive internal combustion engines to increase power density. The broad operating range of the engine also requires a wide range of the turbocharger compressor. At low mass flows, however, turbo compressor operation becomes unstable and eventually enters surge. Surge is characterized by large oscillations in mass flow and pressure. Due to the associated noise, control problems, and possibility of mechanical component damage, this has to be avoided. Different indicators exist to classify compressor operation as stable or unstable on a gas stand. They are based on pressure oscillations, speed oscillations, or inlet temperature increase. In this thesis, a new stability indicator is proposed based on the Hurst exponent of the pressure signal. The Hurst exponent is a number between zero and one that describes what kind of long-term correlations are present in a time series. Data from three cold gas stand experiments are analyzed using this criterion. Results show that the Hurst exponent of the compressor outlet pressure signal has good characteristics. Stable operation is being indicated by values larger than 0.5. As compressor operation moves towards the surge line, the Hurst exponent decreases towards zero. An additional distinction between the long-term correlations of small and large amplitude fluctuations by means of higher order Hurst exponents can be used as an early warning indicator. Further tests using compressor housing accelerometers show that the Hurst exponent is not a good choice for real-time surge detection on the engine. Reasons are the long required sampling time compared to competing methods, and the fact that other periodically repeating oscillations lead to Hurst exponents close to zero independent of compressor operation. / Turboladdare används ofta på förbränningsmotorer för att öka motorns effekttäthet. Motorns breda driftområde ställer krav på ett brett driftområde för turboladdarens kompressor. Vid låga massflöden blir kompressordriften dock mindre stabil, och surge kan uppträda. Surge innebär stora oscillationer i tryck och massflöde genom kompressorn. På grund av oljud, reglerproblem och risken för mekaniska skador vill man undvika surge. Det finns indikatorer för att bedöma kompressorns stabilitet på ett gas stand. Indikatorerna är baserade på tryckoscillationer, varvtalsoscillationer, eller temperaturökning i gasen i kompressorinloppet. I denna avhandling presenteras en ny indikator baserad på Hurst-exponenten, beräknad på trycksignalen. Hurst-exponenten är ett tal mellan noll och ett som beskriver vilka typer av långtidskorrelationer det finns i signalen. Mätningar från tre gas-stand-experiment har analyserats på detta sätt. Analyserna visar att Hurst-exponenten baserad på kompressorutloppstrycket fungerar bra som som surgeindikator. Stabil drift av kompressorn indikeras av att Hurst-exponenten är större än 0.5. När kompressordriftpunkten närmar sig surgelinjen faller Hurst-exponenten mot noll. En distinktion mellan oscillationer med små och stora amplituder kan används för att få en tidig varning. Analyser av vibrationsmätningar på kompressorhuset vid motorapplikation visar att Hurst-exponenten inte är lämplig som realtidsindikator på en motor. Detta kommer sig dels av att data behöver samlas in under en längre tid än med andra tänkbara indikatorer, dels av att andra periodiska oscillationer i signalen kopplade till motorns naturliga beteende leder till Hurst-exponenter nära noll även vid stabil kompressordrift. / <p>QC 20170510</p> / CCGEx - Compressor off-Design

The aim of this thesis is to provide an empirical evidence of multifractality in financial time series and to discuss the relevance of this concept for the current financial theory. We have applied two methods, the Multifractal Detrended Fluctuation analysis and the Generalized Hurst exponent method, on components of the Dow Jones Industrial Average. We analyzed daily data of 30 companies traded on U.S. stock markets from 2002 to 2012. We present results supporting presence of multiscaling in open-close returns. Contrary to published literature, we were not able to find any significant multiscaling in volatility. Moreover based on our analysis, multiscaling is not present in standardized returns and as multifractality requires relatively complicated models, this is our most valuable result. 1

Over the last decades, multifractality has become a downright stylized fact in financial markets. However, its presence has not been adequately statistically proved. The main aim of this thesis is to contribute to the discussion by an ex- tensive statistical analysis of the problem. We investigate returns and volatility of the collection of the four stock indices employing the three popular methods: the GHE, the MF-DFA, and the MF-DMA method. By comparing the results of the original series to those for simulated monofractal series, we conclude that stock market returns as well as volatility exhibit a multifractal nature. Additionally, in order to understand the origin of underlying multifractality, we study vari- ous surrogate series. We found that a fat-tailed distribution significantly affects multifractality. On the other, we were not able to confirm the impact of time correlations as the results strongly depend on the applied model. JEL Classification F12, G02, G10, C12, C22, C49, C58 Keywords econophysics, multifractality, financial markets, Hurst exponent Author's e-mail jana.sedlarikova@gmail.com Supervisor's e-mail kristoufek@ies-prague.org

Made available in DSpace on 2015-04-22T22:07:32Z (GMT). No. of bitstreams: 1 Edilson de Carvalho.pdf: 3411203 bytes, checksum: 3680ef9a40a81ca8cea6def9615ac55d (MD5) Previous issue date: 2012-12-10 / Conselho Nacional de Desenvolvimento Científico e Tecnológico / In this work we analyzed, on the multifractal perspective, the rainfall records from sixteen meteorological stations located in Brazil, especially in the Amazônia, in the towns of Altamira, Araguatins, Cáceres, Corumba, Cruzeiro do Sul, Guaíra, Ibotirama, Manaus, Oriximiná, Piranhas, Porto Velho, Santa Terezinha de Goiás, Santarém, São Paulo de Olivença, Tucuruí and Xambioá. As well the the records of the sea surface temperature anomalies-SSTA for seven regions located in the Atlantic and Pacific oceans, named North Atlantic, South Atlantic and Tropical Atlantic, in the Atlantic ocean, and Nino 1 + 2, Nino 3, Nino 4 and Nino 3:4, in the Pacific ocean. Using the MF-DFA methodology with the addition of the step zero, we calculated the multifractal spectra of the time series related to the rainfall and SSTA records. Also we obtained the correlation between the rainfall and the SSTA data, finding a weak correlation between these series. Based on the Multiplicative Multinomial d-Process, we simulated the zeros in the rainfall series, using the parameters obtained through the polynomial adjustments on the multifractal spectra. Keywords: Time series; rainfall; Multifractality; Hurst exponent / Analisamos neste trabalho, sobre a perspectiva multifractal, os registros de chuva de dezesseis estações meteorológicas localizadas no Brasil e em especial na Amazônia, situadas nas cidades de Altamira, Araguatins, Cáceres, Corumba, Cruzeiro do Sul, Guaíra, Ibotirama, Manaus, Oriximiná, Piranhas, Porto Velho, Santa Terezinha de Goiás, Santarém, São Paulo de Olivença, Tucuruí e Xambioá. Examinamos também, registros de anomalias de temperatura na superfície do mar (SSTA) relativos a sete regiões localizadas no oceanos Atlântico e Pacífico denominadas de Atlântico Norte, Atlântico Sul e Atlântico Tropical no oceano Atlântico e as regiões Nino 1 + 2, Nino 3, Nino 4 e Nino 3:4. Calculamos os espectros multifractais das séries temporais estudadas, referentes aos dados de chuva e de SSTA, utilizando a metodologia MF-DFA com inclusão do passo zero. Medimos os coeficientes de correlação entre os dados de chuva em relação aos dados de SSTA, encontrando uma fraca correlação entre as séries. Com base no d-Processo Multiplicativo Multinomial simulamos zeros em séries de chuva por meio de parâmetros obtidos através de ajustes polinomiais dos espectros multifractais

A necessidade de antecipar e identificar variações de acontecimentos apontam para uma nova direção nos mercados de bolsa de valores e vem de encontro às análises das oscilações de preços de ativos financeiros. Esta necessidade leva a argumentar sobre novas alternativas na predição de séries temporais financeiras utilizando métodos de aprendizado de máquinas e vários modelos têm sido desenvolvidos para efetuar a análise e a previsão de dados de ativos financeiros. Este trabalho tem por objetivo propor o desenvolvimento de um modelo de previsão adaptativo baseado em um sistema SVR-wavelet híbrido, que integra modelos de wavelets e Support Vector Regression (SVR) na previsão de séries financeiras. O método consiste na utilização da Transformada de Wavelet Discreta (DWT) a fim de decompor dados de séries de ativos financeiros que são utilizados como variáveis de entrada do SVR com o objetivo de prever dados futuros de ativos financeiros. O modelo proposto é aplicado a um conjunto de ativos financeiros do tipo Foreign Exchange Market (FOREX), Mercado Global de Câmbio, obtidos a partir de uma base de conhecimento público. As séries são ajustadas gerando-se novas predições das séries originais, que são comparadas com outros modelos tradicionais tais como o modelo Autorregressivo Integrado de Médias Móveis (ARIMA), o modelo Autorregressivo Fracionário Integrado de Médias Móveis (ARFIMA), o modelo Autorregressivo Condicional com Heterocedasticidade Generalizado (GARCH) e o modelo SVR tradicional com Kernel. Além disso, realizam-se testes de normalidade e de raiz unitária para distribuição não linear, tal como testes de correlação, para constatar que as séries temporais FOREX são adequadas para a comprovação do modelo híbrido SVR-wavelet e posterior comparação com modelos tradicionais. Verifica-se também a aderência ao Expoente de Hurst por meio da estatística de Reescalonamento (R/S). / The necessity to anticipate and identify changes in events points to a new direction in the stock exchange market and reaches the analysis of the oscillations of prices of financial assets. This necessity leads to an argument about new alternatives in the prediction of financial time series using machine learning methods. Several models have been developed to perform the analysis and prediction of financial asset data. This thesis aims to propose the development of SVR-wavelet model, an adaptive and hybrid prediction model, which integrates wavelet models and Support Vector Regression (SVR), for prediction of Financial Time Series, particularly Foreign Exchange Market (FOREX), obtained from a public knowledge base. The method consists of using the Discrete Wavelets Transform (DWT) to decompose data from FOREX time series, that are used as SVR input variables to predict new data. The series are adjusted by generating new predictions of the original series, which are compared with other traditional models such as the Autoregressive Integrated Moving Average model (ARIMA), the Autoregressive Fractionally Integrated Moving Average model (ARFIMA), the Generalized Autoregressive Conditional Heteroskedasticity model (GARCH) and the traditional SVR model with Kernel. In addition, normality and unit root tests for non-linear distribution, and correlation tests, are performed to verify that the FOREX time series are adequate for the verification of SVR-wavelet hybrid model and comparison with traditional models. There is also the adherence to the Hurst Exponent through the statistical Rescaled Range (R/S).

Traditional models of financial asset yields are based on a number of simplifying assumptions. Among these are the primary assumptions that changes in asset yields are independent, and that the distribution of these yields is approximately normal. The development of financial asset pricing models has also incorporated these assumptions. A general feature of the pricing models is that the relationship between the model variables is fundamentally linear. Recent empirical research has however identified the possibility for these relations to be non-linear. The empirical research focused primarily on methodological issues relating to the application of the classical rescaled adjusted range. Some of the major issues investigated were: the use of overlapping versus contiguous subseries lengths in the calculation of the statistic's Hurst exponent; the asymptotic distribution of the Hurst exponent for Gaussian time-series and long-term dependent fBm's; matters pertaining to the estimation of the expected rescaled adjusted range. Empirical research in this thesis also considered alternate applications of rescaled range analysis, other than modelling non-linear long-term dependence. Issues relating to the use of the technique for estimating long-term dependent ARFIMA processes, and some implications of long-term dependence for financial time-series have both been investigated. Overall, the general shape of the asymptotic distribution of the Hurst exponent has been shown to be invariant to the level of dependence in the underlying series. While the rescaled adjusted range is a biased indicator of the level of long-term dependence in simulated time-series, it was found that the bias could be efficiently modelled. For real time-series containing structured short-term dependence, the bias was shown to be inconsistent with the simulated results. / Doctor of Philosophy (PhD)

Abstract The impenetrable movement and crash of the stock market is always the most intriguing research task of any financial researcher. Nowadays, it has been proved that the movements of financial asset have the property of non-linearity or near-chaos and shows some tendency within a given period. We used the R/S analysis as the tool to indicate the tendency, and those stocks as our researching objects. We then combined purely price technical analysis indicators and genetic algorithms to form a predicting model. Then we compared our genetic predicting model with the traditional ARIMA analysis and hope to find out the invisible pattern under price volatility. And we hope our model could assist investors in assessing the stock markets more objectively and reduce the risk of stock investment. The researching target is TSMC(2330). We covered the period from 5 September 1994 to 28 December 1999, resulting in 1490 trading days. Historical data are available from Taiwan Economic Journal (TEJ). We execute the researching comparison by bear-market, bull-market, and bull-then-bear market and concluded as follows. 1. After the R/S analysis, we got the Hurst exponent of TSMC to be 0.849855 and the trending cycle was 940. It has proved that the market has tendency and indirectly showed that the Taiwan stock market was not efficient. 2. According to directional precision, our predicting model apparently outpaced the ARIMA model in these three periods. The reason was that our model grabbed more information than the ARIMA model. 3. If we only think about the inputs and outputs, our model seems to be a proper framework for explaining the relationships among variables in comparison with the neural network model having the same input and output variables. 4. We can deduce the invisible relationships of price technical indicators and the closing price. 5. Genetic predicting model can detect the prevailing trend of the learning periods. 6. The shorter the learning period, the better the predicting effects. As a whole and conservatively speaking, we have 70% confidence in directional precision. 7. If we combine proper trading strategy with genetic predicting model and deduct the transaction cost, we still get a better profit than buy-and-hold strategy and have some maneuvering flexibility. 8. After hypothesis testing, our predicting model seems to have some potential of ex ante prediction, but the stability and usability still need further study. In short, we proposed the ex post stock price movement learning model and the viable direction of ex ante prediction. Investors can take advantage of the flexibility of the predicting model and avoid using the over-complex and rigid trading strategies.