This dissertation addresses the general problem of recognition of acoustic signals which may be derived from speech, sonar, or acoustic phenomena. The specific problem of recognizing speech is the main focus of this research. The intention is to design a recognition system for a definite number of discrete words. For this purpose specifically, eight isolated words from the T1MIT database are selected. Four medium length words "greasy," "dark," "wash," and "water" are used. In addition, four short words are considered "she," "had," "in," and "all." The recognition system addresses the following issues: filtering or preprocessing, training, and decision-making. The preprocessing phase uses linear predictive coding of order 12. Following the filtering process, a vector quantization method is used to further reduce the input data and generate a finite inductive sequence of symbols representative of each input signal. The sequences generated by the vector quantization process of the same word are factored, and a single ruling or reference template is generated and stored in a codebook. This system introduces a new modeling technique which relies heavily on the basic concept that all finite sequences are finitely inductive. This technique is used in the training stage. In order to accommodate the variabilities in speech, the training is performed casualty, and a large number of training speakers is used from eight different dialect regions. Hence, a speaker independent recognition system is realized. The matching process compares the incoming speech with each of the templates stored, and a closeness ration is computed. A ratio table is generated anH the matching word that corresponds to the smallest ratio (i.e. indicating that the ruling has removed most of the symbols) is selected. Promising results were obtained for isolated words, and the recognition rates ranged between 50% and 100%.

Creatinine is a metabolic waste product, removed from the blood by the kidneys, and excreted in the urine. The measurement of creatinine is used in the assessment and monitoring of many medical conditions as well as in the determination or adjustment of absorbed dosage of pesticides. Earlier models to predict 24-hour urinary creatinine used ordinary least squares regression and assumed that the subjects' observations were uncorrelated. However, many of these studies had repeated creatinine measurements for each of their subjects. Repeated measures on the same subject frequently are correlated. Using data from the NIOSH-CDC "Pesticide Dose Monitoring in Turf Applicators" study, this thesis project built a model to predict 24-hour urinary creatinine using the Mixed Model methodology. A covariance structure, that permitted multiple observations for any one individual to be correlated, was identified and utilized. The predictive capabilities of this model were then compared to the earlier models investigated.

Doctor of Philosophy / Department of Diagnostic Medicine/Pathobiology / Robert L. Larson / Bradley J. White / Utilization of data analytics allows for rapid and real-time decision making in the food animal production industry. The objective of my research was to implement and utilize different data analytic strategies in multiple sectors of the beef cattle industry in order to determine management, health, and performance strategies. A retrospective analysis using reproductive and genomic records demonstrated that a bull will sire a larger number of calves in a multiple sire-pasture compared to other bulls in the same pasture. A further study was performed to determine if behavior differences existed among bulls in a multiple-sire pasture, and the ability of accelerometers to predict breeding behaviors. Machine learning techniques used classifiers on accelerometer data to predict behavior events lying, standing, walking, and mounting. The classifiers were able to accurately predict lying and standing, but walking and mounting resulted in a lower predictable accuracy due to the extremely low prevalence of these behaviors. Finally, a new form of meta-analysis to the veterinary literature, a mixed treatment comparison, was able to accurately identify differences in metaphylactic antimicrobials on outcomes of bovine respiratory disease morbidity, mortality, and retreatment morbidity. The meta-analysis was not successful in determining the effects of metaphylactic antimicrobials on performance outcomes.

This thesis introduces a novel method for accurate pitch detection and speech segmentation, named Multi-feature, Autocorrelation (ACR) and Wavelet Technique (MAWT). MAWT uses feature extraction, and ACR applied on Linear Predictive Coding (LPC) residuals, with a wavelet-based refinement step. MAWT opens the way for a unique approach to modeling: although speech is divided into segments, the success of voicing decisions is not crucial. Experiments demonstrate the superiority of MAWT in pitch period detection accuracy over existing methods, and illustrate its advantages for speech segmentation. These advantages are more pronounced for gain-varying and transitional speech, and under noisy conditions.

Nous étudions le problème de l'estimation de moyenne et de la densité prédictive d'une population sélectionnée, en obtenant de nouveaux développements qui incluent l'analyse de biais, la décomposition du risque et les problèmes avec restrictions sur les paramètres (chapitre 2). Nous proposons des estimateurs de densité prédictive efficaces en termes de pertes Kullback-Leibler et Hellinger (chapitre 3) améliorant les procédures de plug-in via une perte duale et via une d'expansion de variance. Enfin, nous présentons les résultats de l'amélioration de l'estimateur du maximum de vraisemblance (EMV) d'une moyenne normale bornée pour une classe de fonctions de perte, y compris la perte normale réfléchie, avec des implications pour l'estimation de densité prédictive. A savoir, nous donnons des conditions sur la perte et la largeur de l'espace paramétrique pour lesquels l'estimateur de Bayes par rapport à la loi a priori uniforme sur la frontière domine la EMV. / Abstract : We study the problem of point estimation and predictive density estimation of the mean of a selected population, obtaining novel developments which include bias analysis, decomposition of risk, and problems with restricted parameters (Chapter 2). We propose efficient predictive density estimators in terms of Kullback-Leibler and Hellinger losses (Chapter 3) improving on plug-in procedures via a dual loss and via a variance expansion scheme. Finally (Chapter 4), we present findings on improving on the maximum likelihood estimator (MLE) of a bounded normal mean under a class of loss functions, including reflected normal loss, with implications for predictive density estimation. Namely, we give conditions on the loss and the width of the parameter space for which the Bayes estimator with respect to the boundary uniform prior dominates the MLE.​

The increasing variety of data mining tools offers a large palette of types and representation formats for predictive models. Managing the models then becomes a big challenge, as well as reusing the models and keeping the consistency of model and data repositories. Sustainable access and quality assessment of these models become limited to researchers. The approach for the Data and Model Governance (DMG) makes easier to process and support complex solutions. In this thesis, contributions are proposed towards ensembles of models with a focus on model representation, comparison and usage. Predictive Toxicology was chosen as an application field to demonstrate the proposed approach to represent predictive models linked to data for DMG. Further analysing methods such as predictive models comparison and predictive models combination for reusing the models from a collection of models were studied. Thus in this thesis, an original structure of the pool of models was proposed to represent predictive toxicology models called Predictive Toxicology Markup Language (PTML). PTML offers a representation scheme for predictive toxicology data and models generated by data mining tools. In this research, the proposed representation offers possibilities to compare models and select the relevant models based on different performance measures using proposed similarity measuring techniques. The relevant models were selected using a proposed cost function which is a composite of performance measures such as Accuracy (Acc), False Negative Rate (FNR) and False Positive Rate (FPR). The cost function will ensure that only quality models be selected as the candidate models for an ensemble. The proposed algorithm for optimisation and combination of Acc, FNR and FPR of ensemble models using double fault measure as the diversity measure improves Acc between 0.01 to 0.30 for all toxicology data sets compared to other ensemble methods such as Bagging, Stacking, Bayes and Boosting. The highest improvements for Acc were for data sets Bee (0.30), Oral Quail (0.13) and Daphnia (0.10). A small improvement (of about 0.01) in Acc was achieved for Dietary Quail and Trout. Important results by combining all the three performance measures are also related to reducing the distance between FNR and FPR for Bee, Daphnia, Oral Quail and Trout data sets for about 0.17 to 0.28. For Dietary Quail data set the improvement was about 0.01 though, but this data set is well known as a difficult learning exercise. For five UCI data sets tested, similar results were achieved with Acc improvement between 0.10 to 0.11, closing more the gaps between FNR and FPR. As a conclusion, the results show that by combining performance measures (Acc, FNR and FPR), as proposed within this thesis, the Acc increased and the distance between FNR and FPR decreased.

A novel continuous low-thrust fuel-efficient model predictive control strategy for multi-satellite formations flying in low earth orbit is presented. State prediction relies on a full nonlinear relative motion model, based on quasi-nonsingular relative orbital elements, including earth oblateness effects and, through state augmentation, differential drag. The optimal control problem is specically designed to incorporate latest theoretical results concerning maneuver optimality in the state-space, yielding to a sensible total delta-V reduction, while assuring feasibility and stability though imposition of a Lyapunov constraint. The controller is particularly suitable for networked architectures since it exploits the predictive strategy and the dynamics knowledge to provide robustness against feedback losses and delays. The Networked MPC is validated through real missions simulation scenarios using a high-fidelity orbital propagator which accounts for high-order geopotential, solar radiation pressure, atmospheric drag and third-body effects.

Abstract The utilization of oral lesions as a screening tool for HIV is not well documented. Attendees at two primary health care facilities (Khutsong and Heidelberg) were assessed to determine the predictive value of group I oral lesions for HIV infection. The objectives were to investigate the: 1) HIV prevalence amongst attendees at PHC facilities, 2) Prevalence of HIV-related oral lesions and 3) Correlation between the oral lesions and the HIV status using the Likelihood Ratio test. Methods: All patients over 12 months of age presenting at the two facilities for a curative care consultation over a one-week period (in April 2005) were included. Consent was obtained by trained counselors who also conducted a brief interview and offered pre-test counseling to patients wishing to know their HIV status. Two calibrated dentists conducted a head, neck and oral examination and administered a rapid saliva HIV test (OraQuick HIV-1/2-Rapid HIV-1/2 Antibody Test). Results: A total of 654 attendees were surveyed in the 2 facilities. There was a 100% response. The mean age of the participants was 34 years (range: 1-94), and the majority (73%) were female. HIV prevalence rates were 34% at Khutsong and 36% at Heidelberg. The HIV prevalence peaked at 46% in the 16-45 age groups. Of the 228 who tested positive for HIV, 121 (53%) patients were diagnosed with 1 or more Group I oral lesion. Oral candidiasis (46%) and oral hairy leukoplakia (19%) were the two most common oral lesions diagnosed in the HIV positive cohort. The positive predictive values and specificity values for multiple lesions ranged between 96% and 100%. Most of the likelihood ratios for multiple lesions were greater than 10 which implied that the patients who presented with these lesions were extremely likely to test positive for HIV. The sensitivity values (1% to 37%) and negative predictive values (66% to 70%) remained relatively low. Conclusion: The HIV prevalence of patients attending PHC facilities was high (34%). Oral lesions are useful markers of HIV-infection and should alert clinicians to the presence of HIV infection. Multiple group I lesions were more predictive of HIV infection compared to single lesions.

A thesis submitted to the Faculty of Engineering, University of the Witwatersrand, Johannesburg, in fulfilment of the requirements for the degree of Doctor of Philosophy Johannesburg, October 1997 / This thesis develops and investigates signal processing models that are useful both for interpreting and implementing certain types of Model Predictive Control. Two types of Model Predictive Control are investigated, namely, techniques based on Internal Model Control and Long Range Predictive Control. [Abbreviated abstract. Open document to view full version].

As abelhas e plantas apresentam diferentes graus de especialização em suas interações. Parceiros mais especialistas frequentemente apresentam uma história evolutiva mútua e sobreposição nas áreas de ocorrência. No entanto, a estrutura espacial dos ambientes nos quais esses grupos se distribuem é caracterizada por padrões complexos e dinâmicos. Para analisar a influência dos fatores que atuam na distribuição de espécies de abelhas e plantas associadas foram utilizadas a modelagem de distribuição de espécies, análise multivariada e ferramentas de sistemas de informações geográficas. Os resultados indicaram que a distribuição de gêneros estritamente associados, como é o caso de Peponapis e Cucúrbita, é influenciada pelo clima das áreas de ocupação, bem como provavelmente, por sua história evolutiva e pela domesticação das abóboras (Cucúrbita). Já os gêneros Krameria e Centris apresentam um padrão mais intrincado de distribuição, uma vez que a interação entre ambos é mais complexa. Centris é um grupo diverso que utiliza recursos florais de outras famílias botânicas além de Krameriaceae, o que provavelmente influencia seus padrões de distribuição. No entanto, os resultados obtidos para Krameria demonstraram de maneira geral, a influência de características climáticas na distribuição do grupo e uma provável dependência maior de Centris em alguns casos. Finalmente, foi também demonstrada a importância da inclusão de dados bióticos no processo da modelagem de distribuição, resultando no aumento da acurácia dos modelos e na alteração da projeção da distribuição para o futuro, considerando-se um cenário de mudança climática. Os resultados foram mais conspícuos quando foram consideradas interações mais estreitas entre espécies de abelhas parasitas e hospedeiras do gênero Bombus, do que entre Centris e Krameria. As técnicas utilizadas, em especial a modelagem de distribuição, representaram uma importante contribuição para a análise efetuada. No entanto, embora crescentemente utilizada, a modelagem de distribuição de espécies demanda técnicas e testes mais robustos para avaliar a acurácia dos modelos gerados. Além disso, um desafio adicional a ser vencido consiste no aumento e melhoria da qualidade dos pontos de ocorrência das espécies, principalmente no Brasil. Faz-se necessário um esforço adicional de coleta, especialmente em algumas áreas específicas, bem como, a conservação e digitalização dos dados das coleções biológicas. Porém, as técnicas utilizadas mostraram um grande potencial a ser explorado em outras análises, envolvendo questões biológicas diferentes, ou outros grupos taxonômicos e camadas de dados. / Bees and plants present different degrees of specialism in their interactions. More specialized partners generally present a mutual evolutionary history and overlap with their occurrence areas. Nevertheless, the spatial structure of environments occupied by them is characterized by complex and dynamic patterns. Species distribution modelling, multivariate analyses and geographical system information tools were used in order to analyze the influence of different factors that act in the geographical distribution of associated bees and plants. Results showed that the geographical distribution of close associated genera, such as Peponapis and Cucúrbita are influenced by the clime of occurrence areas, and also, by their evolutionary history and cucurbits domestication (squashes and pumpkins). On the other hand, Centris and Krameria genera presented a more intricate distribution pattern, since their interaction is more complex. Centris is a diverse group that uses other floral resources than those provided by the Krameriacea family, which has probably influenced its distribution, also. However, the results obtained for Krameria showed the influence of clime in its distribution and a stronger relationship with Centris in some cases. Finally, the importance of including biotic data in the species distribution modelling process was also demonstrated, resulting in a general increase in the models accuracy and also altering future scenarios projection, considering climate changes. Stronger interaction, such as the host-parasite bee species of Bombus showed more conspicuous results than those found for Krameria and Centris. The techniques, especially distribution modelling, made an important contribution to the analyses. However, in spite of being increasingly used, distribution modelling demands more robust tests and techniques to evaluate the accuracy of final models. Besides, an additional challenge to be achieved consists in the increase and improvement of species occurrence data, mainly in Brazil. An additional survey effort is necessary, especially in specific areas, as well as the conservation and data digitalization of biological collections. However, the techniques used here showed a great potential to be further explored in other analyses, involving different biological issues, other taxonomic groups and other data layers.