Clustering time-series data into discrete groups can improve prediction as well as providing insight into the nature of underlying, unobservable states of the system. However, temporal heterogeneity and autocorrelation (persistence) in group occupancy can obscure such signals. We use latent-state and hidden Markov models (HMMs), two standard clustering techniques, to model high-resolution hourly movement data from Florida panthers. Allowing for temporal heterogeneity in transition probabilities, a straightforward but rarely explored model extension, resolves previous HMM modeling issues and clarifies the behavioural patterns of panthers. / Thesis / Master of Science (MSc)

No description available.

The objective of this thesis is to design and implement a speech inputting system for addresses in Taiwan,Mainland china and HongKong,The completed system has the capability to identify full census and posting addresses in Taiwan and full posting addresses in Peking¡BShanghai¡BTien-Jin and Chungchin of China¡CFor HongKong,a partial address system,including region/street name or school,hotal and other public location names,is implemented¡C In this thesis,Mel frequency cepstrum coefficient,Hidden Mavkov model and lexicon search strategy are applied to choose the initial address candidates¡FMandarin intonation classification technique is then used to increase the final correct rate,under speaker dependent case,a 90%correct rate can be reached by using a Intel Celeron 2.4GHz CPU and RedHat Linux 9.0 operating system¡CThe total address-inputting task can be completed within 3 seconds¡C

This thesis develops new hidden Markov models and applies them to financial market and macroeconomic time series. Chapter 1 proposes a probabilistic model of the return distribution with rich and heterogeneous intra-regime dynamics. It focuses on the characteristics and dynamics of bear market rallies and bull market corrections, including, for example, the probability of transition from a bear market rally into a bull market versus back to the primary bear state. A Bayesian estimation approach accounts for parameter and regime uncertainty and provides probability statements regarding future regimes and returns. A Value-at-Risk example illustrates the economic value of our approach. Chapter 2 develops a new efficient approach to model and forecast time series data with an unknown number of change-points. The key is assuming a conjugate prior for the time-varying parameters which characterize each regime and treating the regime duration as a state variable. Conditional on this prior and the time-invariant parameters, the predictive density and the posterior of the change-points have closed forms. The conjugate prior is further modeled as hierarchical to exploit the information across regimes. This framework allows breaks in the variance, the regression coefficients or both. In addition to the time-invariant structural change probability, one extension assumes the regime duration has a Poisson distribution. A new Markov Chain Monte Carlo sampler draws the parameters from the posterior distribution efficiently. The model is applied to Canadian inflation time series. Chapter 3 proposes an infinite dimension Markov switching model to accommodate regime switching and structural break dynamics or a combination of both in a Bayesian framework. Two parallel hierarchical structures, one governing the transition probabilities and another governing the parameters of the conditional data density, keep the model parsimonious and improve forecasts. This nonparametric approach allows for regime persistence and estimates the number of states automatically. A global identification algorithm for structural changes versus regime switching is presented. Applications to U.S. real interest rates and inflation compare the new model to existing parametric alternatives. Besides identifying episodes of regime switching and structural breaks, the hierarchical distribution governing the parameters of the conditional data density provides significant gains to forecasting precision.

This thesis develops new hidden Markov models and applies them to financial market and macroeconomic time series. Chapter 1 proposes a probabilistic model of the return distribution with rich and heterogeneous intra-regime dynamics. It focuses on the characteristics and dynamics of bear market rallies and bull market corrections, including, for example, the probability of transition from a bear market rally into a bull market versus back to the primary bear state. A Bayesian estimation approach accounts for parameter and regime uncertainty and provides probability statements regarding future regimes and returns. A Value-at-Risk example illustrates the economic value of our approach. Chapter 2 develops a new efficient approach to model and forecast time series data with an unknown number of change-points. The key is assuming a conjugate prior for the time-varying parameters which characterize each regime and treating the regime duration as a state variable. Conditional on this prior and the time-invariant parameters, the predictive density and the posterior of the change-points have closed forms. The conjugate prior is further modeled as hierarchical to exploit the information across regimes. This framework allows breaks in the variance, the regression coefficients or both. In addition to the time-invariant structural change probability, one extension assumes the regime duration has a Poisson distribution. A new Markov Chain Monte Carlo sampler draws the parameters from the posterior distribution efficiently. The model is applied to Canadian inflation time series. Chapter 3 proposes an infinite dimension Markov switching model to accommodate regime switching and structural break dynamics or a combination of both in a Bayesian framework. Two parallel hierarchical structures, one governing the transition probabilities and another governing the parameters of the conditional data density, keep the model parsimonious and improve forecasts. This nonparametric approach allows for regime persistence and estimates the number of states automatically. A global identification algorithm for structural changes versus regime switching is presented. Applications to U.S. real interest rates and inflation compare the new model to existing parametric alternatives. Besides identifying episodes of regime switching and structural breaks, the hierarchical distribution governing the parameters of the conditional data density provides significant gains to forecasting precision.

The objective of this thesis is to design and implement a speech recognition system for Mandarin names and phrases. This system utilizes Mel frequency cepstral coefficients, hidden Markov model and lexicon search strategy to select the phrase candidates. The experimental results indicate that for the speaker dependent case, a strategy incorporating overlapping frames and hybrid training can result in an improvement of 4%, 5%, 4% and 2% on the recognition rate for the Mandarin name, two-word, three-word and four-word phrase recognition systems respectively. Under Redhat Linux 9.0 operating system, any Mandarin name or phrase can be recognized within 2 seconds by a computer with Intel Celeron 2.4 GHz CPU.

Dissertation (PhD)--University of Stellenbosch, 2007. / Bibliography. Also available via the Internet.

In this thesis, we investigate two di erent approaches to animal movement modelling; nite mixture models, and di usion processes. These models are considered in two di erent contexts, rstly for analysis of data obtained in home range studies, and then, on a much smaller scale, modelling the movements of larvae. We consider the application of mixture models to home range movement data, and compare their performance with kernel density estimators commonly used for this purpose. Mixtures of bivariate normal distributions and bivariate t distributions are considered, and the latter are found to be good models for simulated and real movement data. The mixtures of bivariate t distributions are shown to provide a robust parametric approach. Subsequently, we investigate several measures of overlap for assessing site delity in home range data. Di usion processes for home range data are considered to model the tracks of animals. In particular, we apply models based on a bivariate Ornstein-Uhlenbeck process to recorded coyote movements. We then study modelling in a di erent application area involving tracks. Di usion models for the movements of larvae are used to investigate their behaviour when exposed to chemical compounds in a scienti c study. We nd that the tted models represent the movements of the larvae well, and correctly distinguish between the behaviour of larvae exposed to attractant and repellent compounds. Mixtures of di usion processes and Hidden Markov models provide more exible alternatives to single di usion processes, and are found to improve upon them considerably. A Hidden Markov model with 4 states is determined to be optimal, with states accounting for directed movement, localized movement and stationary observations. Models incorporating higherorder dependence are investigated, but are found to be less e ective than the use of multiple states for modelling the larvae movements.

No description available.

Precipitation is an important element in the hydrological cycle. To predict and simulate large-scale precipitation, Global Circulation Models (GCMs) are widely used. However, their grid scale is too big to apply to local hydrologic fields. In this study, non-homogenous hidden Markov models (NHMM) are explored as a means of generating the probability of precipitation occurrence in small scale given large-scaled weather patterns. Three different spatial models: (1) independent (2) auto-logistic (3) Chow-Liu tree, are also explored, along with methods and steps for parameter estimation. From this exploration, independent models with NHMM are recommended for very small precipitation networks, and the maximum likelihood method is found to be the most practical fitting method. If there are many points for downscaling, Chow-Liu tree models with the Expectation-Maximization (EM) algorithm are recommended. If more exact solutions are needed, auto-logistic models can be employed. If many points are considered in auto-logistic models, the (EM) algorithm should be used to estimate parameters separately and global optimization methods should be used for emission matrix. The major problem found with the NHMM model in this study is matching the rainfall amount for each year or month. This problem can be addressed by whether combining occurrence models with amount modes or by improving only occurrence models.