Thèses sur le sujet « Index switching »

In this experiment I investigated the relationship between set-switching and transfer learning, both of which presumably invoke executive functioning (EF), which may in turn be correlated with intelligence. Set-switching was measured by a computerized version of the Wisconsin Card Sort Task. Another computer task was written to measure learning-transfer ability. The data indicate little correlation between the ability to transfer learning and the capacity for set-switching. That is, these abilities may draw from independent cognitive mechanisms. The major difference may be requirement to utilize previous learning in a new way in the learning-transfer task.

Nelle tecnologie di sequenziamento di nuova generazione, centinaia o migliaia di campioni di DNA possono essere sequenziati simultaneamente (multiplexing) e le letture di sequenziamento ottenute possono essere distinte dalla presenza di sequenze nucleotidiche specifiche del campione (indici) incorporate nei primer utilizzati per l'amplificazione del DNA. Pipeline bioinformatiche personalizzate, leggendo gli indici presenti nelle letture di sequenziamento li assegnano ad uno specifico campione (demultiplexing). Il multiplexing, tuttavia, è afflitto dal cambio di indice, un fenomeno che si verifica quando i primer di indice libero vengono fusi in modo casuale a sequenze di DNA appartenenti ad altri campioni non correlati del pool di librerie e determinano l'assegnazione errata delle sequenze a uno o più campioni errati. Nel campo della terapia genica (GT), gli studi sui siti di integrazione dei vettori (IS) dipendono fortemente dal sequenziamento di frammenti di DNA (contenenti giunzioni genomiche provirali-cellulari) da diversi campioni e sono influenzati dallo scambio dell'indice. Questo problema è particolarmente rilevante negli studi di tracciamento clonale, dove il livello di IS condiviso tra diverse linee cellulari o diversi punti temporali dello stesso paziente con GT è richiesto per definire i livelli di ricostituzione multilineare e stimare il numero di cellule staminali e altri calcoli. Pertanto, la diffusione dell'IS tra i set di dati causata dalla scambio dell'indice potrebbe comportare livelli di condivisione IS non veritieri che potrebbero portare a un'interpretazione errata dei risultati. Per valutare l'entità della scambio dell'indice nelle analisi IS, abbiamo analizzato 123.431.269 letture di sequenziamento provenienti da un pool composto da 54 campioni amplificati in triplice copia, ciascuno contrassegnato da due indici fusi alle estremità dei prodotti PCR contenenti LTR e Linker Sequenze di cassette (LC) risultanti in 162 combinazioni di indici (che combinano un totale di 48 indici LTR e 32 LC). Da questa analisi abbiamo scoperto che> 95% delle letture di sequenziamento apparteneva alle 162 combinazioni di indici corrette mentre il restante 5% di letture appartiene a 1374 combinazioni di falsi indici risultanti da eventi franchi di scambio di indici. I livelli di scambio erano simili tra i diversi indici LTR e LC con una media di 1709 ± 3469 letture (intervallo da 9 a 52000) per le combinazioni di falsi indici. Abbiamo quindi valutato i livelli di condivisione di IS mappato in modo univoco tra diversi campioni e abbiamo scoperto che essenzialmente tutti i campioni avevano diversi livelli di contaminazione. Complessivamente, il 91,5% di IS è stato assegnato a un singolo campione, il 7,25% è stato trovato condiviso in due campioni e il restante 1,25% era presente in più di 2 campioni. Concentrandoci su un campione di una linea cellulare con 6 IS noti, abbiamo calcolato i livelli di diffusione e la loro abbondanza relativa su altri campioni. Da questa analisi abbiamo scoperto che almeno uno dei 6 noti IS è stato trovato in 13 campioni non correlati su 54 (24%). In 3 campioni su 13 la quantità di letture contaminanti da questa linea cellulare ha raggiunto livelli che vanno dal 13 al 40% dell'intero set di dati. Questi livelli elevati di contaminazioni hanno giustificato lo sviluppo di nuovi approcci per l'indicizzazione della correzione della scambio negli studi IS. A questo scopo abbiamo sviluppato una serie di algoritmi probabilistici e logici che consente di rimuovere sequenze contaminanti. Questo studio è iniziato con l'analisi del sito di integrazione, ma successivamente è stato esteso ad altri campi diversi. Nella tesi viene mostrato un nuovo metodo per pulire i dataset da questo tipo di contaminazioni.
In Next generation sequencing technologies, hundreds or thousands of DNA samples can be sequenced simultaneously (multiplexing) and the obtained sequencing reads can be distinguished by the presence of sample-specific nucleotide sequences (indexes) embedded in the primers used for the DNA amplification. Custom bioinformatics pipelines, by reading the indexes present in the sequencing reads assign them to a specific sample (demultiplexing). Multiplexing however is plagued by index switching, a phenomenon occurring when free index primers are randomly fused to DNA sequences belonging to other unrelated samples of the library pool and resulting in the incorrect assignment of sequences to one or multiple wrong samples. In the field of gene therapy (GT) (see Appendix A), vector integration site (IS) studies heavily depend on sequencing of DNA fragments (containing proviral-cellular genome junctions) from several samples and are affected by index switching. This issue is particularly relevant in clonal tracking studies, where the level of shared IS between different cell lineages or different time points of the same GT patient are required to define the levels of multilineage reconstitution and estimate the number of stem cells and other calculations. Therefore, the spreading of IS between datasets caused by index switching could result in inflated sharing IS levels which could lead to misinterpretation of the results. To evaluate the extent of index switching in IS analyses, we analysed 123,431,269 sequencing reads originating from a pool composed by 54 samples amplified in triplicate, each tagged by two indexes fused to the ends of the PCR products containing the LTR and Linker Cassette (LC) sequences resulting in 162 index combinations (combining a total of 48 LTR and 32 LC indexes). From this analysis we found that >95% of sequencing reads belonged to the correct 162 index combinations while a the remaining 5% of reads belonging to 1374 false index combinations resulting from frank events of index swapping. The levels of swapping were similar among the different LTR and LC indexes with an average of 1709 ± 3469 reads (range 9 to 52000) for false index combinations. We then evaluated the levels of sharing of univocally mapped IS between different samples and found that essentially all samples had different levels of contamination. Overall, 91.5% of IS were assigned to a single sample, 7.25% were found shared in two samples and the remaining 1.25% were present in more than 2 samples. Focusing on a sample from a cell line with 6 knowns IS we calculated the spreading levels and their relative abundance on other samples. From this analysis we found that at least one of the 6 know IS were found in 13 unrelated samples out of 54 (24%). In 3 out of 13 samples the amount of contaminating reads from this cell line reached levels ranging from 13 to 40% of the entire dataset. These high levels of contaminations justified the development of new approaches for indexing switching correction in IS studies. To this aim we developed a set of probabilistic and logic algorithms that allows to remove contaminating sequences. This study started with the integration site analysis, but after was extended to other different fields. In the thesis is showed a new method for cleaning dataset from this kind of contaminations.

Demonstration of ultrafast all-optical signal processing (AOSP) using silicon as the active material has been limited by large two-photon absorption loss and long lifetimes of the resulting free carriers. For AOSP at speeds in the terahertz, an order of magnitude faster than that the fastest current electronic counterpart, a class of π-conjugated organic molecules called polymethines provides a promising alternative to silicon as they possess large third-order nonlinearities, and ultrafast polarization response to an incident field. The challenge in the application of polymethines as active nonlinear optical materials for AOSP is in translating their promising molecular properties into bulk material properties. The large linear polarizability and charged nature of the polymethines molecules strongly promote aggregation and phase-separation in solid blends, offsetting their advantageous molecular optical properties. In this work, polymethines’ resistance to deleterious spontaneous symmetry breaking and aggregation was enhanced by substitutions of metal- and chalcogen- containing terminal groups, and rigid steric groups above and below the π-conjugated plane of polymethine chain. The resulting polymethines/amorphous polycarbonate (APC) blend films demonstrated an unprecedentedly high two-photon figure-of-merit, |Re(χ(3))/Im(χ(3))| and low linear loss. The optical quality of the polymethines/APC films was also improved by replacing the commonly-used alkyl ammonium counterions with more polarizable aryl phosphonium counterions with moderate ground state dipole moment. The resulting dye-polymer blend films showed an enhanced near-infrared transparency while its magnitude of the third-order susceptibility, |χ(3)|, showed a good agreement with that extrapolated from the molecular third-order polarizability, γ. For facile integration of these promising organic materials into SOH, the substrate surface was functionalized using silane coupling chemistry for the reduction of surface energy mismatch between the polymer films and the waveguide containing substrates. The optical and SEM micrographs showed vastly improved coverage and infiltration of the microfeatures. Furthermore, to enable the precise engineering of waveguide cross-sectional dimensions for single-mode propagation in the organic cladding, the dispersion curves of the polymethines/polymer blends were generated using prism coupling and ellipsometry. The combined efforts in the development of molecules and materials discussed in the thesis have culminated into a successful identification and optimization of the polymethines dyes and their polymer blends for imminent demonstrations of on-chip AOSP at terahertz speed.

L’Europe a subi la crise la plus sévère de sa récente histoire à la suite de la crise financière globale de 2008. C’est pourquoi cette thèse a l’objectif d’identifier de façon empirique les déterminants de cette crise dans le cadre de 15 principaux membres de l’UE. Dans ce sens, nous développons d’abord un index de pression fiscale continu, contrairement aux travaux empiriques précédents, afin d’identifier des périodes de crise dans les pays UE-15 de 2003 à 2015. Ensuite, nous utilisons trois différentes techniques d’estimation, à savoir Cartes auto-organisatrices, Logit et Markov. Nos résultats d’estimation démontrent que notre indicateur de crise identifie le timing et la durée de la crise de dette dans chacun des pays de UE-15. Résultats empiriques indiquent également que l’occurrence de la crise de dette dans l’UE-15 est la conséquence de la détérioration de balances macroéconomiques et financières sachant que les variables comme le ratio des prêts non-performants sur les crédits totaux du secteur bancaire, la croissance du PIB, chômage, balance primaire / PIB, le solde ajusté du cycle PIB. De plus, variables démontrant la qualité de gouvernance tel que participation et responsabilisation, qualité de la réglementation, et de l'efficacité gouvernementale, jouent également un rôle important dans l’occurrence et sur la durée de la crise de dette dans le cadre de l’UE-15. Étant donne que les résultats économétriques indiquent l’importance de la détérioration fiscale dans l’occurrence de la crise de dette européenne, nous testons la convergence fiscale des pays membre de l’UE. Les résultats montrent que Portugal, Irlande, Italie, Grèce et Espagne diverge des autres pays de l’UE-15 en termes de dette publique / PIB alors qu’ils convergent, à part la Grèce, avec les autres pays membres de l’UE-15 en termes de déficit budgétaires / PIB
Europe went through the most severe economic crisis of its recent history following the global financial crisis of 2008. Hence, this thesis aims to empirically identify the determinants of this crisis within the framework of 15 core EU member countries (EU-15). To do so, the study develops a continuous fiscal stress index, contrary to previous empirical studies that tend to use event-based crisis indicators, which identifies the debt crises in the EU-15 and the study employs three different estimation techniques, namely Self-Organizing Map, Multivariate Logit and Panel Markov Regime Switching models. Our estimation results show first that the study identifies correctly the time and the length of the debt crisis in each EU-15-member country by developing a fiscal stress index. Empirical results also indicate, via three different models, that the debt crisis in the EU-15 is the consequence of deterioration of both financial and macroeconomic variables such as nonperforming loans over total loans, GDP growth, unemployment rates, primary balance over GDP, and cyclically adjusted balance over GDP. Besides, variables measuring governance quality, such as voice and accountability, regulatory quality, and government effectiveness, also play a significant role in the emergence and the duration of the debt crisis in the EU-15. As the econometric results clearly indicate the importance of fiscal deterioration on the occurrence of the European debt crisis, this study also aims to test the fiscal convergence among the EU member countries. The results indicate that Portugal, Ireland, Italy, Greece, and Spain diverge from other EU-15 countries in terms of public debt-to-GDP ratio. In addition, results also show that all PIIGS countries except for Greece converge to EU-10 in terms of budget deficit-to-GDP ratio

Investors constantly seek information that provides an edge over the market. One of the conventional methods is to find factors which can predict asset returns. In this study we improve the Fama and French Five-Factor model with Regime-Switches, student's t distributions and copula dependencies. We also add price momentum as a sixth factor and add a one-day lag to the factors. The Regime-Switches are obtained from a Hidden Markov Model with conditional Student's t distributions. For the return process we use factor data as input, Student's t distributed residuals, and Student's t copula dependencies. To fit the copulas, we develop a novel approach based on the Expectation-Maximisation algorithm. The results are promising as the quantiles for most of the portfolios show a good fit to the theoretical quantiles. Using a sophisticated Stochastic Programming model, we back-test the predictive power over a 26 year period out-of-sample. Furthermore we analyse the performance of different factors during different market regimes.

碩士
國立清華大學
工業工程與工程管理學系
103
Acceptance sampling plans are practical tools for quality management applications and provide the producer and consumer a general rule for lot sentencing and reduce the cost. As the rapid advancement of manufacturing technology, suppliers required their products to be of high quality with very low fraction of defectives often measured in Parts Per Million（PPM）. Process capability indices are convenient and powerful tools for measuring process performance and it is widely used in Acceptance sampling plans. This paper attempts to develop a Variables Quick Switching Sampling System（VQSS System）based on the Cpk index. The VQSS System consists of two inspection plan along with a set of switching rules between them. The first sampling plan, called normal inspection plan, is applied for periods with good quality. The second sampling plan, called tightened inspection plan, is applied for use during problems encountered periods. The probabilities of acceptance under normal inspection and tightened inspection are derived, and the proposed VQSS System is also developed based on the exact sampling distribution rather than approximation approach. Three types of the proposed plan parameters are determined by solving a non-linear optimization problem with two-point conditions on the OC curve. The behavior and performance of the proposed sampling system is discussed and also compared with the conventional variables sampling plans. Finally, tables of the plan parameters for various selected quality levels and risks are provided for practical applications.

碩士
國立清華大學
工業工程與工程管理學系
104
Acceptance sampling plan is a practical quality tool which evaluates only parts of products and then decides whether to accept or not on the submitted lot. Quick switching sampling (QSS) system is combined with two variables single sampling plans under normal inspection and tightened inspection. 　　In order to provide numerical measures on process performance, several process capability indices (PCIs) have been applied. Taguchi capability index is developed by incorporating the Taguchi loss function, and takes the process targeting and process variability into consideration simultaneously. 　　This research develops a variables QSS system of three types based on Taguchi capability index. To determine the parameters of three types of system, the problems are formulated as optimization programming. The objective function is to minimize the average sample number (ASN), and the constraints functions are set to fulfill two-point conditions on the operating characteristic (OC) curve. The performance of proposed system is discussed and compared with single sampling plan. Lastly, an application example is illustrated.

碩士
國立臺灣海洋大學
應用經濟研究所
98
The ordinary least squares (OLS) technique (Ederington, 1979; Figlewski, 1984), the co-integration method (Ghosh, 1993; Lien and Luo, 1993), and the bivariate GARCH-type models allowing time-varying nature in asset returns (Baillie and Myers, 1991; Kroner and Sultan, 1993; Park and Switzer, 1995; Gagnon and Lypny, 1995; Kavussanos and Nomikos, 2000; Bystrom, 2003) are the most common approaches to estimate minimum-variance hedge ratios. However, those conventional approaches have been used to calculate the optimal hedge ratios in a sense of linear correlation which could result in bias estimates if the joint distribution of spot and futures is not elliptical and/or is non-linear. Since copula functions of asymmetric dependence structures and extreme values can capture the extreme co-movements of spot and futures, this study builds a dependence-switching model (DS model), which is integrated by copula functions and Markov-switching model by Hamilton (1989, 1994) and is allowed that the dependence of spot and futures can switch between two different structures. We construct a hedging portfolio via the DS model and evaluate the dynamic hedging performance. The results show that the DS model outperforms the conventional approaches such as OLS, ECM, and DCC-GARCH.

碩士
國立政治大學
金融研究所
104
This paper divided into two parts to investigate on the information content of CBOE SKEW Index. For the first part, we do time series analysis to observe the relationship between SKEW Index and other variables. First, we found that SKEW index is totally different from VIX index. VIX index is a proxy for the standard deviation of the returns. The standard deviation describes the average spread of the distribution of returns around its mean. This is not a sufficient measure of risk because the distribution of S&P 500 log returns is not normal. SKEW Index captures the tail risk of the distribution. Next, SKEW Index is good at predict future S&P500 ETF returns especially weekly speaking. Also, we found that the correlation between SKEW index & S&P500 index is too unstable to interpret. We argue that it’s not easy to interpret SKEW Index directly but we can combine SKEW Index with VIX Index. Regarding the above reason, in second part, we combined SKEW Index with VIX Index to construct trading strategy under Markov Switching Model. By comparing with FTP Model, which included VIX index only, we found that TVTP model, which encompassed VIX Index and SKEW Index together, significantly outperform others. When the model detected regime switching, we buy/short SPY ETF in the market separately. We did the simulation test from 2002.4.15 to 2013.3.29. Without considering tax, fee and dividend, we earned yearly average rate of return 13.61%. After considering tax, fee and dividend, we earned yearly average rate of return 9.51%.

碩士
淡江大學
管理科學學系碩士班
104
Stock and futures markets are the two main markets in Taiwan’s financial market. With the world trend of financial liberalization and internationalization, the relations between the spot market and the future market have been becoming a hot topic of in-vestors. Therefore, the research subjects in this study are the Taiwan Stock Exchange Capitalization Weighted Stock Index (TAIEX) and Taiwan Index Futures (TX1) from July 21, 1998 to December 31, 2015. The dynamic influence of price jumps of index futures on spot prices with the regime-switching model proposed by Hamilton (1989) is explored. Being compared to other researched models, the SWARCH model proposed by Hamilton and Susmel (1994) can best capture the price behavior of Taiwan Stock Mar-ket on the likelihood ratio test. The resultant showed that futures price volatility has a positive impact on spot prices and futures price volatility has no significant impact on spot price volatility. Besides, the probability of price volatility persistence in Taiwan stock market is very high and price fluctuations may be susceptible to global economic and political factors.

Dissertação apresentada como requisito parcial para obtenção do grau de Mestre em Estatística e Gestão de Informação.
There is an increasing recognition among marketing scholars and practitioners of the importance of the influence of switching costs and attractiveness of alternatives in the relationship between customer satisfaction and customer loyalty. To date, however, there is a lack of research about the process through which these variables influence the satisfaction-loyalty relationship. This dissertation aims to evaluate the importance of switching costs and attractiveness of alternatives in explaining customer loyalty. Using a revised model of the European Customer Satisfaction Index (ECSI), applied to the banking industry, this study intends to include switching costs as perceived by customers and the attractiveness of alternatives as independent antecedents of customer loyalty and as moderators of the impact of satisfaction on loyalty. Both direct and moderating effects of switching costs and attractiveness of alternatives are tested, using a methodology based on structural equation models. The main findings of this study indicate that both constructs influence loyalty directly and the strength of the satisfaction-loyalty relationship.

As Multi-Protocol Label Switched (MPLS) networks increase in usage and size, the number of traffic engineered tunnels or Label Switched Paths (LSPs) which must be established has an impact on network state maintenance, administration and scalability. The ability to signal and meet Quality-of-Service (QoS) requirements in such networks has been addressed through the addition of Differentiated Services (Diff-Serv) mappings and other traffic engineering mechanisms. However, for the purpose of path computation, route advertisements, signaling and admission control, multiple traffic classes carried together are still treated as a single class. This work explores extensions to MPLS which allow for the accommodation of up to eight distinct traffic classes per label switched path. Through an examination of simulation results. a comparison between existing methods and the proposed additions is made that shows scenarios in which such traffic class aggregation or "bundling" provides a significant reduction in the number of paths which must be maintained in the network.

碩士
國立政治大學
統計研究所
99
Black and Scholes (1973) proposed B-S model to fit asset return, but B-S model can’t effectively explain some asset return properties, such as leptokurtic, volatility smile, volatility clustering and long memory. Merton (1976) develop jump diffusion model (JDM) that consider abnormal information of market will affect the stock price, and this model can explain leptokurtic and volatility smile of asset return at the same time. Charles, Fuh and Lin (2011) extended the JDM and proposed regime-switching jump independent model (RSJIM) that consider jump rate is related to market states. RSJIM not only retains JDM properties but describes volatility clustering and long memory. In this paper, we extend RSJIM to regime-switching jump dependent model (RSJDM) which consider jump size and jump rate are both related to market states. We use EM and SEM algorithm to estimate parameters and covariance matrix, and use LR test to compare RSJIM and RSJDM. By using 1999 to 2010 Dow-Jones industrial average index and S&P 500 index as empirical evidence, RSJDM can explain index return properties said before. Finally, we calculate index option price formulation by Esscher transformation and do sensitivity analysis and market validation which give the smallest error of option prices by RSJDM.

碩士
華梵大學
工業工程與經營資訊學系碩士班
107
Sampling inspection plan plays a very important role on quality control, which provides both buyer and seller a decision method of lot sentencing, extensively applied in the inspection of raw materials, semi-products, finished-products and outgoing. In this research, we develop two types of quick switching sampling system based on coefficient of variation (CV), and the two types of systems are composed of normal inspection and tighten inspection. Type I system means the acceptance value of normal inspection differs from that of tighten inspection while their sample size is the same. Type II system means the acceptance value of normal inspection is the same to that of tighten inspection while their sample size is different. The sample size and critical acceptance values of the proposed plan are determined by minimizing the sample size under the condition of satisfying the two-points of operating characteristic (OC) curve. For practical purpose, the parameters of the proposed plan are provided for some combinations of quality levels with commonly used producer's risk and consumer's risk. In addition, the proposed plan is compared with the existing sampling inspection plans based on CV in terms of sample size and OC curve. Research results shows that type I system has a better performance that has a smaller sample size significantly than those of other sampling inspection plans, which can save lots of inspection costs.

碩士
國立虎尾科技大學
經營管理研究所
97
The main concept of asset allocation stems from the portfolio theory of Markowitz (1952), which is the average - variance portfolio optimization theory. However, in the face of the business periodic economy, there must be a different asset allocation strategy to reduce risk and increase the return rate effectively. The subject of this study will delve into Taiwan 50 Index Constituents for the portfolio return, and utilize Markov-Regime Switching Model to predict the future of the constituent stocks under the different transition probability of the expected return and variance. Using the market model to obtain the residual stock return time series, and then utilizes an average of variance to obtain the optimal weight of various types of assets to predict the best allocation of the investment portfolio risk. The assumption of this study is bull market and bear market. On the basis of these two portfolios, we give those two different weights, and make use of traditional Sharpe Index to compare the portfolio performance. Empirical research results indicate as follows: （1）In this study, the result of Markov-Regime Switching Model on Asset Allocation found out that the standard deviation of the accumulated reward can be significant beyond the traditional model of capital assets. Therefore, the Sharpe ratio performance surpasses not only traditional model of capital assets but also Taiwan 50 Index ETF asset allocation. （2）With the practical experiment of this study, we found out that we can take advantage of the index of industrial production dividing the state of de facto rather than constituent stocks of the average rate of return. However, using different indicators will result different assets and affects final Sharpe ratio performance, which performs will be better than Taiwan 50 Index ETF Asset Allocation.

碩士
真理大學
管理科學研究所
92
In this paper, Gray’s(1996) generalized regime-switching(GRS) model and autoregressive distributed lag(ADL) model are used to capture the return and volatility dynamics between the Taiwan stock index and futures markets. Using GRS model to capture the conditional expectation and conditional variance series, we find that the model that allows both the conditional expectation and conditional variance to change with two regimes is an appropriate model. In the low-volatility regime, the return of stock index and futures both are positive and its volatility have GARCH effect. In the high-volatility regime, the return of stock index and futures both are negative, but only the futures has a GARCH process in volatility behavior. In addition, the expected duration of the low-volatility regime is longer than that of the high-volatility regime. Using ADL model to study the return and volatility dynamics between the Taiwan stock index and futures markets. The return of stock index leads temporarily to itself and futures. The volatility of futures leads temporarily to itself and stock index, but the stock index does not have cross-market volatility spillovers effect. There is a long-run equilibrium relationship between the return of stock index and futures, but there is no long-run equilibrium relationship between the volatility of stock index and futures.

碩士
國立政治大學
統計研究所
104
To model asset return, Chen, Chang, Wen and Lin (2013) proposed Markov-Modulated Jump Diffusion Model (MMJDM) assuming that the Brownian motion term and jump frequency are all related to market states. In fact, the interest rate is not constant, Regime-Switching Model is taken to fit the process of the zero-coupon bond price, and a bivariate model for interest rate and stock index return with regime-switching and dependent jump risks (MMJDMSI) is proposed. The empirical data are Dow Jones Industrial Average and S&;P 500 Index from 1999 to 2013, together with US 1-Year Treasury Bond over the same period. Model parameters are estimated by the Expectation-Maximization (EM) algorithm. The likelihood ratio test (LRT) is performed to compare nested models, and MMJDMSI is better than the others. Then, European call option pricing formula under each model is derived via Esscher transformation, and sensitivity analysis is conducted to evaluate changes resulted from different parameter values under the MMJDMSI pricing formula. Finally, model calibrations are performed and implied volatilities are computed under each model empirically. In cases of in-the-money and out-the-money, MMJDMSI has either the smallest or the second smallest pricing error. Also, the implied volatilities from MMJDMSI display a volatility smile curve.

碩士
國立成功大學
財務金融研究所
96
This study is to investigate the effectiveness of various price and volatility forecast models including Error Correction EGARCH model (EC-EGARCH), Regime-Switching EGARCH model (RS-EGARCH), and hybrid models by combining Genetic Algorithm (GA) with EC-EGARCH (EC-EGARCHGA) and RS-EGARCH (RS-EGARCHGA). Since EGARCH rather than GARCH considers information asymmetry, EGARCH is used as conditional variance model in this study. Hybrid models by combining EC-EGARCH and RS-EGARCH with Genetic Algorithm (GA) are used in this study to see whether one of artificial intelligence, GA, can help improve forecast ability in price level and volatility since it has been supported by Neely and Weller (2002) in predicting the volatility and Lai and Li (2006) in predicting earning per share. In this thesis, 10-minute S&P 500 index and index futures are used and the sample period spreads from January 1986 to December 2007. Because yearly 10-minut interval data is hard to converge when EC-EGARCH or RS-EGARCH model is used, we only use the fourth quarter data in each year. Based on Mean Absolute Percentage Error (MAPE), Mean Absolute Error (MAE), Paired t Test and Fisher Exact Test, we conclude that two hybrid models, EC-EGARCHGA and RS-EGARCHGA, perform better relative to their corresponding ordinary econometric models in price and volatility forecasts.

Gene networks can be modeled by piecewise-linear (PL) switching systems of differential equations, called Glass networks after their originator. Networks of interacting genes that regulate each other may have complicated interactions. From a `systems biology' point of view, it would be useful to know what types of dynamical behavior are possible for certain classes of network interaction structure. A useful way to describe the activity of this network symbolically is to represent it as a directed graph on a hypercube of dimension $n$ where $n$ is the number of elements in the network. Our work here is considering this problem backwards, i.e. we consider different types of cycles on the $n$-cube and show that there exist parameters, consistent with the directed graph on the hypercube, such that a periodic orbit exists. For any simple cycle on the $n$-cube with a non-branching vertex, we prove by construction that it is possible to have a stable periodic orbit passing through the corresponding orthants for some sets of focal points $F$ in Glass networks. When the simple cycle on the $n$-cube doesn't have a non-branching vertex, a structural principle is given to determine whether it is possible to have a periodic orbit for some focal points. Using a similar construction idea, we prove that for self-intersecting cycles where the vertices revisited on the cycle are not adjacent, there exist Glass networks which have a periodic orbit passing through the corresponding orthants of the cycle. For figure-8 patterns with more than one common vertex, we obtain results on the form of the return map (Poincar{\'e} map) with respect to how the images of the returning cones of the 2 component cycle intersect the returning cone themselves. Some of these allow complex behaviors.

碩士
國立成功大學
財務金融研究所
97
This study contains two parts. First of all, the study investigates the effectiveness of various price forecasting models including Error Correction EGARCH model (EC-EGARCH), regime-switching EGARCH model (RS-EGARCH), hybrid model by combining EC-EGARCH with Genetic Algorithms (EC-EGARCHGA) and hybrid model by using combining RS-EGARCH with Genetic Algorithms (RS-EGARCHGA). Second, we use forward simulation to examine whether investors can earn positive excess return net of transaction cost based on predictive value about S&P500 index futures and SPDR index fund by Error Correction EGARCH model (EC-EGARCH), regime-switching EGARCH model (RS-EGARCH). Since EC-EGARCH model considers both conditional mean and variance at the same time, it can help us predict price more completely. As to the RS-EGARCH model, it relies on different coefficients in each regime to account for the possibility that the financial series may undergo a finite numbers of changes over the sample period. In addition, Genetic Algorithm (GA), is also used in this study to see whether artificial intelligence, GA, can help improve forecasting ability in price level since it has been supported by Lai and Li (2006) in predicting earning per share. Based on Mean Absolute Percentage Error (MAPE), Mean Absolute Error (MAE), Paired t Test and Fisher Exact Test, we can evaluate which chosen models, EC-EGARCH, RS-EGARCH, EC-EGARCHGA and RS-EGARCHGA, performs best in predicting price. Furthermore, we examine whether the investors can earn significant excess return by Z Test.

In this thesis, an extensive simulation study is done to gauge the performance of parallel access to multiple mirror sites on the Internet. The study is based on the Digital Fountain approach designed by J. Byers et al. in which Tornado codes are used to minimize decoding time at the expense of injecting extra packets into the system. In this study, both Reed-Solomon and Tornado codes are considered. The results indicate that randomly permuting the packets at every mirror site is not the optimal solution. A new technique, which staggers the packets, is shown to be superior. This study employs OPNET Modeler, which is a powerful event driven simulation tool. Simulations results show that using Reed-Solomon codes with staggered packet transmission provides dramatically improved system performance.

碩士
國立臺北大學
合作經濟學系
94
A Study on the Relationships , Asymmetric Volatility Switching and Mean Reverting Property for Stock Price Index, Exchange Rate and Foreign Capitals in Taiwan：An Application of Multivariate VAR ANST GARCH-M Model ABSTRACT: The purposes of this study is to explore relationships among stock price index , exchange rate and foreign capitals in Taiwan and to detect whether these markets exist asymmetric volatility switching and asymmetric mean reverting property or not .To achieve this purpose in this research, the multivariate GARCH models which include the asymmetric nonlinear smooth transition（anst）GARCH-M procedure are applied to obtain empirical evidence. The results of this research are shown as follows： 1. In conditional mean equations ,besides VAR VS GARCH monthly model, the empirical evidences of other models have shown, the change of the foreign capitals leads to the stock price index and exchange rate to change. It implies that foreign capitals may be thought as the role of price discovery. 2. In daily data, the evidences of VAR VS GARCH model have shown all of these three conditional variances have the asymmetric volatility switching effects on them. In the situation of monthly data, the volatility behavior of exchange rate has the same effects on it, except stock price index and exchange rate. 3. Base on the empirical evidences of VAR ANST GARCH-M model, it indicates that all of these three conditional means have asymmetric mean reverting behavior, and it implies that foreign capitals has overreactions in daily data. 4. In the assessment of the establish model, in this research it shows that the VAR ANST GARCH-M model have the best forecasting ability . Keyword：Stock Price Index, Exchange Rate, Foreign Capitals, ANST GARCH, Overreaction

We present a model of coupled transcriptional-translational ultradian oscillators (TTOs) as a possible mechanism for the circadian rhythm observed at the cellular level. It includes nonstationary Poisson interactions between the transcriptional proteins and their affined gene sites. The associated reaction-rate equations are nonlinear ordinary differential equations of stochastic switching type. We compute the deterministic limit of this system, or the limit as the number of gene-proteins interactions per unit of time becomes large. In this limit, the random variables of the model are simply replaced by their limiting expected value. We derive the Kolmogorov equations — a set of partial differential equations —, and we obtain the associated moment equations for a simple instance of the model. In the stationary case, the Kolmogorov equations are linear and the moment equations are a closed set of equations. In the nonstationary case, the Kolmogorov equations are nonlinear and the moment equations are an open-ended set of equations. In both cases, the deterministic limit of the moment equations is in agreement with the deterministic state equations.

This research study focuses on the application models of Data Mining and Machine Learning covering cellular network traffic, in the objective to arm Mobile Network Operators with full view of performance branches (Services, Device, Subscribers). The purpose is to optimize and minimize the time to detect service and subscriber patterns behaviour. Different data mining techniques and predictive algorithms have been applied on real cellular network datasets to uncover different data usage patterns using specific Key Performance Indicators (KPIs) and Key Quality Indicators (KQI). The following tools will be used to develop the concept: RStudio for Machine Learning and process visualization, Apache Spark, SparkSQL for data and big data processing and clicData for service Visualization. Two use cases have been studied during this research. In the first study, the process of Data and predictive Analytics are fully applied in the field of Telecommunications to efficiently address users’ experience, in the goal of increasing customer loyalty and decreasing churn or customer attrition. Using real cellular network transactions, prediction analytics are used to predict customers who are likely to churn, which can result in revenue loss. Prediction algorithms and models including Classification Tree, Random Forest, Neural Networks and Gradient boosting have been used with an exploratory Data Analysis, determining relationship between predicting variables. The data is segmented in to two, a training set to train the model and a testing set to test the model. The evaluation of the best performing model is based on the prediction accuracy, sensitivity, specificity and the Confusion Matrix on the test set. The second use case analyses Service Quality Management using modern data mining techniques and the advantages of in-memory big data processing with Apache Spark and SparkSQL to save cost on tool investment; thus, a low-cost Service Quality Management model is proposed and analyzed. With increase in Smart phone adoption, access to mobile internet services, applications such as streaming, interactive chats require a certain service level to ensure customer satisfaction. As a result, an SQM framework is developed with Service Quality Index (SQI) and Key Performance Index (KPI). The research concludes with recommendations and future studies around modern technology applications in Telecommunications including Internet of Things (IoT), Cloud and recommender systems.
Cellular networks have evolved and are still evolving, from traditional GSM (Global System for Mobile Communication) Circuit switched which only supported voice services and extremely low data rate, to LTE all Packet networks accommodating high speed data used for various service applications such as video streaming, video conferencing, heavy torrent download; and for say in a near future the roll-out of the Fifth generation (5G) cellular networks, intended to support complex technologies such as IoT (Internet of Things), High Definition video streaming and projected to cater massive amount of data. With high demand on network services and easy access to mobile phones, billions of transactions are performed by subscribers. The transactions appear in the form of SMSs, Handovers, voice calls, web browsing activities, video and audio streaming, heavy downloads and uploads. Nevertheless, the stormy growth in data traffic and the high requirements of new services introduce bigger challenges to Mobile Network Operators (NMOs) in analysing the big data traffic flowing in the network. Therefore, Quality of Service (QoS) and Quality of Experience (QoE) turn in to a challenge. Inefficiency in mining, analysing data and applying predictive intelligence on network traffic can produce high rate of unhappy customers or subscribers, loss on revenue and negative services’ perspective. Researchers and Service Providers are investing in Data mining, Machine Learning and AI (Artificial Intelligence) methods to manage services and experience. This research study focuses on the application models of Data Mining and Machine Learning covering network traffic, in the objective to arm Mobile Network Operators with full view of performance branches (Services, Device, Subscribers). The purpose is to optimize and minimize the time to detect service and subscriber patterns behaviour. Different data mining techniques and predictive algorithms will be applied on cellular network datasets to uncover different data usage patterns using specific Key Performance Indicators (KPIs) and Key Quality Indicators (KQI). The following tools will be used to develop the concept: R-Studio for Machine Learning, Apache Spark, SparkSQL for data processing and clicData for Visualization.
Electrical and Mining Engineering
M. Tech (Electrical Engineering)