Sur le marché mondial de la gestion collective, la France occupe la deuxième place en termes du nombre de fonds et du montant d’actifs nets gérés. Malgré le rôle important des OPCVM obligataires dans la diversification des portefeuilles, ils sont néanmoins peu étudiés par rapport aux fonds actions. Notre recherche étudie la performance des fonds obligataires français sur une période de 20 ans, entre 1988 et 2007. Elle s’articule sur quatre chapitres. Le premier présente le panorama du développement des différentes catégories de fonds obligataires et leur couple risque / rentabilité.Le chapitre 2 construit un échantillon de 81 fonds grâces aux données généreusement fournies par la société Morningstar France. Après avoir testé l’homogénéité des distributions de rentabilités des fonds obligataires, on constate que les sous-catégories d’un même horizon de placement ont une distribution statistiquement identique. Ce qui permet de regrouper les fonds en trois catégories selon leur maturité : Long Terme, Moyen Terme et Court Terme. Le chapitre suivant se penche sur la non normalité des distributions de rentabilités (avec le test Jarque-Bera) et la recherche de l’origine de ce phénomène. Il y a absence de l’effet du mois de janvier ; les rentabilités sont non stationnaires et auto-corrélées. Les deux méthodes de correction pour auto-corrélation des rentabilités entrainent une augmentation du risque sans modifier la rentabilité moyenne ; mais ne modifient pas le classement des fonds. Le dernier chapitre est consacré à l’évaluation de la performance des fonds obligataires.Les douze mesures les plus utilisées par les professionnels conduisent à des classements des fonds quasiment identiques et à se demander si le ratio de Sharpe n’est pas finalement la seule mesure à utiliser. Sans aller jusque-là, la méthode de la « classification hiérarchique ascendante » (empruntée à la technique du marketing) permet de réduire à quatre les mesures de performance sans perte d’informations sur le classement des fonds entre eux. Ce chapitre confirme également la présence de la persistance de la performance à l’instar de certaines études anglo-saxonnes sur cette question. / At the world level, the French sector of mutual funds is the second for the number of fundsand the amount of net assets managed. Curiously enough regarding its size, the French bond fundshave been « forgotten » by the researchers and the present thesis seems to be the first one on thissubject. It studies their performances over a 20-year period (from January 1988 to December 2007)with the help of four chapters. The first one is a survey of that sector development and measures thereturn/risk ratio for the different kinds of bond funds. Chapter two presents the 81-funds samplegraciously given by Morningstar France. The analysis of the returns distributions shows that thedifferent funds with and without income distribution with the same maturity have the same statisticaldistribution; it enables to group the funds according to their maturity, namely Long Term, Middle Termand Short Term. Chapter three analyses the non-normality of distributions with the Jarque-Bera testand tries to find the origin of this phenomenon. There is no January effect but returns are nonstationaryand auto-correlated. The two correcting methods for that auto-correlation lead to anincrease of the measured risk without changing the average return and the funds rankings are notmodified. The last chapter deals with the performance measurement of these bond funds. The twelvemeasures most used by professionals give almost the same funds rankings. One can wonder if theSharpe ratio is not the unique relevant measure to use. A technique borrowed from the marketingscience called «cluster analysis» enables to reduce the twelve measures to only four without losinginformation on the funds rankings. This chapter also shows that the performance is persistent, resultalso found by some American researches on this question.

Introdução: A Organização Mundial da Saúde sugere que as recomendações alimentares para populações devem basear-se em alimentos ao invés de nutrientes. No entanto, devem levar em consideração, que os indivíduos não consomem nutrientes ou alimentos isoladamente e sim refeições compostas por uma variedade de alimentos. Objetivo: Identificar os padrões alimentares do café da manhã, almoço e jantar da população adulta e idosa do município de São Paulo e caracterizá-los de acordo com variáveis socioeconômicas, demográficas e antropométricas. Métodos: Foram utilizados dados secundários do estudo transversal de base populacional ISA-Capital 2008 de adultos e idosos (n=1102). O consumo alimentar foi estimado pelo Multiple Source Method considerando dois recordatórios de 24 horas. A partir dos grupos de alimentos de cada refeição aplicou-se análise fatorial por componentes principais (rotação varimax) para derivar os padrões alimentares. Calcularam-se os escores fatoriais, que foram então utilizados para agrupar os indivíduos através da análise de agrupamento. Adicionou-se um grupo com os indivíduos que não realizaram a refeição para cada uma das análises. Os grupos foram caracterizados segundo sexo, faixa etária, renda familiar per capita e estado nutricional. Admitiu-se nível de significância de 0,05 e para caracterização dos grupos considerou-se uma diferença de escore padronizado em pelo menos |0,5| com relação à amostra geral. Resultados: Identificou-se prevalência de 5,6 por cento de omissão do café da manhã, 3,6 por cento do almoço e 12,8 por cento do jantar. Quanto aos que realizaram as refeições, identificaram-se três padrões do Café da Manhã: Saudável, Tradicional e Lanche; quatro padrões do Almoço: Tradicional, Salada, Suco Adoçado e Ocidental; e quatro padrões do Jantar: Café com Leite e Pão, Transição, Tradicional e Sopa e Frutas. Os grupos de indivíduos associados a idosos apresentaram aderência ao café da manhã Saudável, almoço Salada e jantar Sopas e Frutas. Aqueles associados a homens e adultos em geral omitiram o café da manhã e aderiram ao almoço Ocidental e jantar Transição. Os grupos associados a apenas homens aderiram a café da manhã e almoço Tradicional. Aqueles associados a somente adultos apresentaram aderência ao almoço Suco Adoçado. E associados a mulheres aderiram a um jantar Café com Leite e Pão. Grupos associados à baixa escolaridade aderiram a um jantar Sopa e Frutas, e os de escolaridade elevada aderiram a café da manhã Lanche e Omissão do mesmo; e padrão de almoço ocidental. Grupos associados com indivíduos de baixa renda aderiram ao almoço tradicional. Conclusão: Os resultados encontrados no presente estudo indicam que padrões alimentares identificados de acordo com a refeição discriminam bem o consumo alimentar, salientando peculiaridades que não são encontradas em análises globais. E a associação entre a aderência a esses padrões e fatores demográficos e socioeconômicos sugere a existência de públicos-alvo para o planejamento e execução de políticas públicas em alimentação e nutrição. / Introduction: The World Health Organization suggests that dietary recommendations for populations should be based on foods rather than nutrients. However, individuals do not consume foods or nutrients in isolation but meals with variety of foods. Objective: To identify dietary patterns from breakfast, lunch and dinner of the adult and elderly population in São Paulo and characterize them according to socioeconomic, demographic and anthropometric variables. Methods: Data from cross-sectional population-based ISA Capital 2008 of adults and elderly (n = 1102) was used. Dietary intake was estimated by Multiple Source Method considering two 24-hour recalls. Principal component factor analysis (varimax rotation) was used to derive dietary patterns for each meal. Factor scores were calculated and then used in the cluster analysis. Individuals that skipped the meal were added as an extra cluster. The groups were characterized by gender, age, family income and nutritional status. Significance level of 0.05 was assumed for the groups characterization it was considered a difference of standardized score of at least | 0.5 | with respect to the overall sample. Results: We found omission prevalence of 5.6 per cent for breakfast, 3.6 per cent for lunch and 12.8 per cent for dinner. Three patterns of breakfast were identified: Healthy, Traditional and Snack; four patterns to lunch: Traditional, Salad, Sweetened Juice and Western; and four patterns to dinner: Coffee with Milk and Bread, Transitional, Traditional and Soup and Fruits. The groups of individuals associated with the elderly showed adherence to Healthy breakfast, Salad lunch and Soups and Fruit dinner. The groups that were associated with adult and men skipped breakfast and adhered to the Western lunch and Transition dinner. The groups associated with only men showed adherence to Traditional breakfast and lunch. The groups associated with only adults showed adherence to Sweetened Juice lunch. The groups associated with women showed adherence with Coffee with Milk and Bread dinner. The groups associated with low education adhered to Soup and Fruits dinner, and the groups associated with high school skipped breakfast and they adhered to Snack breakfast and Western breakfast. Groups associated with low-income individuals adhered to Traditional lunch. Conclusion: The results of this study indicate that dietary patterns identified by meal discriminated food consumption very well, highlighting peculiarities that a global analysis cannot show. And the adherence to meal patterns associated with socioeconomic and demographic factors suggest there is evidence of risk groups and possible public policies related to health diet.

<p>As pedagogical innovations continue to be developed and adopted in engineering education, it is important to understand how these innovations affect the students’ experiences and achievements. A common data analysis practice when evaluating educational innovations is to aggregate the data from all of the students together. However, this data aggregation inherently biases the results toward the characteristics of the dominant student group, leaving the experiences of minority groups largely unexplored. In this dissertation, I investigate the students’ experiences and achievements in an undergraduate dynamics course, and I intentionally use analysis methods that disaggregate the data to better understand the behaviors and performance of smaller subgroups of students, not just the majority.</p> <p> This dissertation presents three studies that examine: 1) the validity, reliability, and fairness of a standardized set of conceptual questions on the final exam, with a focus on gender fairness, 2) how and why the students use the available resources, and 3) how the students’ holistic resource usage patterns relate to their academic achievement. My motivation for choosing these studies was that conceptual assessments and customized resources are two key components of the learning environment for the dynamics course. However, the quality of the conceptual exam questions used for the course had yet to be evaluated. Similarly, the learning environment for the course incorporates many customized resources, including a custom-written “lecturebook” (a hybrid of a textbook and a workbook) and an extensive online library of videos, but little was known about how the students used these resources, or how the students’ pattern of resource usage related to their performance in the course. </p> <p> The first study in this dissertation used multiple-group confirmatory factor analysis to investigate item-level gender bias in a 12-item Abbreviated Dynamics Concept Inventory (aDCI), which was a set of standardized conceptual questions included on the final exam. The results suggested that two items were slightly biased against women, with stereotypically-masculine contexts and content as possible sources of the bias. The bias in the aDCI items likely unfairly lowered some women’s final exam scores, highlighting the need for engineering educators to consider the fairness of their assessments.</p> <p> The second study used a cluster analysis of survey responses to identify nine archetypical patterns of resource usage, all of which differed from the average resource-usage pattern of the aggregated sample. An analysis of forty-four student interviews, organized by resource-usage cluster, determined that students exhibited their resource-usage behaviors largely because of how they perceived the resource’s availability, accessibility, and quality. The results illustrate that there is no “typical” way in which the students used the resources, so it is important for instructors to consider a wide array of usage behaviors when designing a course’s learning environment and resources.</p> <p> The third study utilized a multiple regression analysis to find that <i>on average</i> a student’s resource-usage pattern is not related to their achievement when controlling for many other demographic, cognitive, and non-cognitive factors that can affect resource usage and performance. However, two individual resource-usage patterns were significantly related to achievement. Students who primarily used their lecturebook and their peers for support performed better than their similar peers in other resource-usage clusters. Conversely, students who rarely used their lecturebook had lower course grades than their peers. Drawing from the results of the second study, general study-habit suggestions for the students in the course were extracted from the qualitative themes found in the interviews of the students in these two clusters.</p> <p> Overall, the results of these three studies highlight how the experiences and achievements of smaller groups of students would go unnoticed if analytical methods that only utilized aggregated data were used. While the setting of this research is specific to the assessments and resources of a given dynamics course, the methods used to disaggregate the data to gain insights about different subgroups of students are applicable to many engineering education contexts. My hope is that this work inspires more researchers to consider the experiences of all students, not just those of the majority.</p>

Lu Yang. / Thesis submitted in: November 2004. / Thesis (M.Phil.)--Chinese University of Hong Kong, 2005. / Includes bibliographical references (leaves 96-103). / Abstracts in English and Chinese. / Abstract --- p.i / Acknowledgement --- p.iv / Chapter 1 --- Introduction --- p.1 / Chapter 1.1 --- Motivation --- p.1 / Chapter 1.2 --- Our Contributions --- p.3 / Chapter 1.3 --- Thesis Outline --- p.5 / Chapter 2 --- Background Study --- p.7 / Chapter 2.1 --- MPEG-7 Introduction --- p.8 / Chapter 2.2 --- Image Analysis in CBIR Systems --- p.11 / Chapter 2.2.1 --- Color Information --- p.13 / Chapter 2.2.2 --- Color Layout --- p.19 / Chapter 2.2.3 --- Texture Information --- p.20 / Chapter 2.2.4 --- Shape Information --- p.24 / Chapter 2.2.5 --- CBIR Systems --- p.26 / Chapter 2.3 --- Image Processing in JPEG Frequency Domain --- p.30 / Chapter 2.4 --- Photo Album Clustering --- p.33 / Chapter 3 --- Feature Extraction and Similarity Analysis --- p.38 / Chapter 3.1 --- Feature Set in Frequency Domain --- p.38 / Chapter 3.1.1 --- JPEG Frequency Data --- p.39 / Chapter 3.1.2 --- Our Feature Set --- p.42 / Chapter 3.2 --- Digital Photo Similarity Analysis --- p.43 / Chapter 3.2.1 --- Energy Histogram --- p.43 / Chapter 3.2.2 --- Photo Distance --- p.45 / Chapter 4 --- 1-Dimensional Photo Album Management Techniques --- p.49 / Chapter 4.1 --- Photo Album Sorting --- p.50 / Chapter 4.2 --- Photo Album Clustering --- p.52 / Chapter 4.3 --- Photo Album Compression --- p.56 / Chapter 4.3.1 --- Variable IBP frames --- p.56 / Chapter 4.3.2 --- Adaptive Search Window --- p.57 / Chapter 4.3.3 --- Compression Flow --- p.59 / Chapter 4.4 --- Experiments and Performance Evaluations --- p.60 / Chapter 5 --- High Dimensional Photo Clustering --- p.67 / Chapter 5.1 --- Traditional Clustering Techniques --- p.67 / Chapter 5.1.1 --- Hierarchical Clustering --- p.68 / Chapter 5.1.2 --- Traditional K-means --- p.71 / Chapter 5.2 --- Multidimensional Scaling --- p.74 / Chapter 5.2.1 --- Introduction --- p.75 / Chapter 5.2.2 --- Classical Scaling --- p.77 / Chapter 5.3 --- Our Interactive MDS-based Clustering --- p.80 / Chapter 5.3.1 --- Principal Coordinates from MDS --- p.81 / Chapter 5.3.2 --- Clustering Scheme --- p.82 / Chapter 5.3.3 --- Layout Scheme --- p.84 / Chapter 5.4 --- Experiments and Results --- p.87 / Chapter 6 --- Conclusions --- p.94 / Bibliography --- p.96

Finite mixture modeling is a useful statistical technique for clustering individuals based on patterns of responses. The fundamental idea of the mixture modeling approach is to assume there are latent clusters of individuals in the population which each generate their own distinct distribution of observations (multivariate or univariate) which are then mixed up together in the full population. Hence, the name mixture comes from the fact that what we observe is a mixture of distributions. The goal of this model-based clustering technique is to identify what the mixture of distributions is so that, given a particular response pattern, individuals can be clustered accordingly. Commonly, finite mixture models, as well as the special case of latent class analysis, are used on data that inherently involve repeated measures. The purpose of this dissertation is to extend the finite mixture model to allow for the number of repeated measures to be incorporated and contribute to the clustering of individuals rather than measures. The dimension of the repeated measures or simply the count of responses is assumed to follow a truncated Poisson distribution and this information can be incorporated into what we call a dimension informative finite mixture model (DIMM). The outline of this dissertation is as follows. Paper 1 is entitled, "Dimension Informative Mixture Modeling (DIMM) for questionnaire data with an informative number of repeated measures." This paper describes the type of data structures considered and introduces the dimension informative mixture model (DIMM). A simulation study is performed to examine how well the DIMM fits the known specified truth. In the first scenario, we specify a mixture of three univariate normal distributions with different means and similar variances with different and similar counts of repeated measurements. We found that the DIMM predicts the true underlying class membership better than the traditional finite mixture model using a predicted value metric score. In the second scenario, we specify a mixture of two univariate normal distributions with the same means and variances with different and similar counts of repeated measurements. We found that that the count-informative finite mixture model predicts the truth much better than the non-informative finite mixture model. Paper 2 is entitled, "Patterns of Physical Activity in the Northern Manhattan Study (NOMAS) Using Multivariate Finite Mixture Modeling (MFMM)." This is a study that applies a multivariate finite mixture modeling approach to examining and elucidating underlying latent clusters of different physical activity profiles based on four dimensions: total frequency of activities, average duration per activity, total energy expenditure and the total count of the number of different activities conducted. We found a five cluster solution to describe the complex patterns of physical activity levels, as measured by fifteen different physical activity items, among a US based elderly cohort. Adding in a class of individuals who were not doing any physical activity, the labels of these six clusters are: no exercise, very inactive, somewhat inactive, slightly under guidelines, meet guidelines and above guidelines. This methodology improves upon previous work which utilized only the total metabolic equivalent (a proxy of energy expenditure) to classify individuals into inactive, active and highly active. Paper 3 is entitled, "Complex Drug Use Patterns and Associated HIV Transmission Risk Behaviors in an Internet Sample of US Men Who Have Sex With Men." This is a study that applies the count-informative information into a latent class analysis on nineteen binary drug items of drugs consumed within the past year before a sexual encounter. In addition to the individual drugs used, the mixture model incorporated a count of the total number of drugs used. We found a six class solution: low drug use, some recreational drug use, nitrite inhalants (poppers) with prescription erectile dysfunction (ED) drug use, poppers with prescription/non-prescription ED drug use and high polydrug use. Compared to participants in the low drug use class, participants in the highest drug use class were 5.5 times more likely to report unprotected anal intercourse (UAI) in their last sexual encounter and approximately 4 times more likely to report a new sexually transmitted infection (STI) in the past year. Younger men were also less likely to report UAI than older men but more likely to report an STI.

Os royalties de petróleo têm ganhado notoriedade nos últimos anos pelo crescimento das atividades petrolíferas no país, as mudanças nas leis, preços do petróleo e transferência e aplicação de recursos nos Municípios para seu desenvolvimento socioeconômico. Esta pesquisa busca identificar mudanças em indicadores socioeconômicos em Municípios do estado do Rio de Janeiro através de análise temporal de variáveis que indiquem uma relação da evolução socioeconômica dos Municípios com os royalties de petróleo. Para testar esta hipótese, foi feita uma análise de agrupamento dos Municípios do estado do Rio de Janeiro utilizando as variáveis IDH (Índice de Desenvolvimento Humano), PIB (Produto Interno Bruto) e população, sendo todas as variáveis apresentando valores do ano 2000. Com a determinação do número de grupos, foram escolhidos dois Municípios de cada grupo sob a condição de maior arrecadador de royalties e não arrecadador de royalties e analisados entre os anos 2003 e 2006 para verificar a possível mudança de indicadores socioeconômicos neste período e possível relação desta mudança com royalties de petróleo. / The oil royalties have gained notoriety in recent years by the growth of oil activities in the country, changes in the legislation, oil prices and the transfer and application of resources in the municipalities for their socioeconomic development. The purpose of this research is to indentify changes in the socioeconomic indicators of municipalities of state of Rio de Janeiro through temporal analysis of variables that indicate a relationship between the socioeconomic development of municipalities and the royalties from oil. To test this hypothesis, we performed a cluster analysis on municipalities of Rio de Janeiro state using variables IHD (Index of Human Development), GDP (Gross Domestic Product) and population, and all the variables are related to the year 2000. Upon determination of the number of groups, the first was the received the largest amount of royalties, and the second did not received royalties at all. They were analyzed from 2003 to 2006 in order to investigate changes in the socioeconomic indicators, and if those changes had to do with oil royalties.

The main goal of this thesis is to test the possibility of forward and backward segmentation on the single-family housing market, to reveal, describe and develop a profile of segments and to suggest marketing recommendations. Data from primary research is encoded in MS Excel and processed in the statistical analysis program SPSS Statistics 19 for Windows. The output of the thesis reveals three forward segments and three backward segments of the market. Based on certain specifics shown by each of the monitored segments, I suggested numerous marketing recommendations.

A leishmaniose visceral é uma zoonose de grande importância para a saúde pública, com ampla distribuição geográfica e epidemiologia complexa. Apesar de diversas estratégias de controle, a doença continua se expandindo, tendo o cão como principal reservatório. Levando em consideração que análises espaciais são úteis para compreender melhor a dinâmica da doença, avaliar fatores de risco e complementar os programas de prevenção e controle, o presente estudo teve como objetivo caracterizar a distribuição da leishmaniose visceral canina e relacionar sua dinâmica com características ou feições espaciais no município de Panorama (SP). A partir de dados secundários coletados em um inquérito sorológico entre agosto de 2012 e janeiro de 2013, 986 cães foram classificados como positivos e negativos de acordo com o protocolo oficial do Ministério da Saúde. Posteriormente uma análise espacial foi conduzida, compreendendo desde a visualização dos dados até a elaboração de um mapa de risco relativo, passando por análises de cluster global (função K) e local (varredura espacial). Para avaliar uma possível relação entre o cluster detectado com a vegetação na área de estudo, calculou-se o Índice de Vegetação por Diferença Normalizada (NDVI). A prevalência da doença encontrada na população de cães estudada foi de 20,3% (200/986). A visualização espacial demonstrou que tanto animais positivos quanto negativos estavam distribuídos por toda a área de estudo. O mapa de intensidade dos animais positivos apontou duas localidades de possíveis clusters, quando comparado ao mapa de intensidade dos animais negativos. As análises de cluster confirmaram a presença de um aglomerado e um cluster foi detectado na região central do município, com um risco relativo de 2,63 (p=0,01). A variação espacial do risco relativo na área de estudo foi mapeada e também identificou a mesma região como área significativa de alto risco (p<0,05). Não foram observadas diferenças no padrão de vegetação comparando as áreas interna e externa ao cluster. Sendo assim, novos estudos devem ser realizados com o intuito de compreender outros fatores de risco que possam ter levado à ocorrência do cluster descrito. A prevalência, a localização do cluster espacial e o mapa de risco relativo fornecem subsídios para direcionamento de esforços do Setor de Vigilância Epidemiológica de Panorama para áreas de alto risco, o que pode poupar recursos e aperfeiçoar o controle da leishmaniose visceral no município. / Visceral leishmaniasis is a zoonosis of great importance for public health, having worldwide distribution and complex epidemiology. Despite several control strategies, the disease continues to expand, being dogs its main reservoir. Considering that spatial analysis is useful to better understand the disease dynamics, to evaluate risk factors and to complement prevention and control programs, this study aimed to characterize the distribution of canine visceral leishmaniasis and relate it to spatial features in Panorama, an endemic city of São Paulo State, Brazil. Using secondary data collected in a serological survey between August 2012 and January 2013, 986 dogs were classified as positive and negative for leishmaniasis. Later, a spatial analysis was conducted comprising from data visualization to the elaboration of a relative risk map, undergoing global (K function) and local cluster analysis (spatial scan). To evaluate a possible relation between the detected cluster and vegetation in the study area, the Normalized Difference Vegetation Index (NDVI) was calculated. The prevalence in the studied dog population was 20.3% (200/986). Spatial visualization showed that both positive and negative animals were distributed throughout the study area. Intensity map of positive animals showed two sites of possible clusters compared to the intensity map of negative animals. A cluster was detected in a central region of the city, with a relative risk of 2.63 (p=0,01). The spatial variation of the relative risk in the study area was mapped and the same region was identified as a significant high risk area (p<0.05). No differences were observed in the vegetation pattern comparing the cluster internal and external areas. Therefore, further studies should be conducted in order to understand other risk factors that may have led to the occurrence of the described cluster. The prevalence data, the cluster location and the relative risk map provide subsidies for the efforts of the city epidemiological surveillance department towards high risk areas. This can save resources and improve control of visceral leishmaniasis in the study area.

O efeito das variáveis meteorológicas na maturação da cana-de-açúcar é um processo ainda pouco conhecido, apesar de apresentar inúmeros impactos à qualidade da matéria-prima para a indústria sucro-energética brasileira. O objetivo do presente trabalho foi avaliar os efeitos das variáveis meteorológicas no processo de maturação de cultivares de cana-de-açúcar e gerar modelos capazes de descrevê-lo. O experimento foi realizado em área pertencente ao Grupo Raízen (ex-Cosan), em Piracicaba, SP, no período de março de 2002 a outubro de 2003. Foram avaliadas oito cultivares de cana-de-açúcar em 32 amostragens realizadas de março a outubro de 2003. Variáveis relacionadas à qualidade de matéria-prima da cana-de-açúcar (ATR, AR, ART, Pureza, Brix, Pol%cana, fibra e umidade) e variáveis meteorológicas (temperatura do ar, precipitação, radiação solar, entre outras) foram submetidas à análise estatística descritiva e multivariada, visando melhorar a compreensão do processo de maturação e suas relações com as condições ambientais. As variáveis meteorológicas de melhor correlação com as variáveis de qualidade da cana-de-açúcar foram utilizadas na elaboração de modelos descritivos do processo de maturação, os quais foram agrupados em precoces, médios e tardios, de acordo com o padrão de maturação das cultivares estudadas. A avaliação dos modelos foi realizada pela análise dos resíduos (presença de outliers, homogeneidade das variâncias e normalidade dos resíduos) e pela comparação com dados independentes (erros médios, R2, índice de concordância de Willmott e índice de confiança de Camargo). Independente do padrão de maturação (precoce, médias e tardias), os modelos de melhor desempenho foram aqueles que descreveram as variáveis ATR, Brix, Pol%cana e umidade, com valores de R2 ajustado acima de 0,9 (p significativo a 1%) e ótimo desempenho segundo índice C de Camargo (acima de 0,85), quando testados com dados independentes. O modelo de fibra apresentou os menores valores de R2 ajustado (em torno de 0,65) e não normalidade dos resíduos. Contudo, sua avaliação com dados independentes levou à obtenção de valores de índice C acima de 0,8. Os modelos gerados mostram que a precipitação pode ser utilizada como a principal variável na previsão da qualidade da matéria-prima da cana-de-açúcar e que seu uso pode ser estendido a outras áreas além daquela para onde foi inicialmente gerado. / The effect of meteorological variables on sugarcane ripening is a not very well known process in spite of the several impacts of that on the quality of the raw material for the sugarcane industry. The objective of this study was to evaluate the effects of meteorological variables on sugarcane ripening and based on that to establish models able to describe this process. The study was conducted in an area of the Raízen Group (former Cosan), in Piracicaba, state of São Paulo, Brazil, from March 2002 (planting) to October 2003. Eight sugarcane cultivars were evaluated by analyzing 32 samples collected from March to October 2003. Variables related to the quality of the raw material (ATR, AR, ART, Purity, Brix, Pol%cane, fiber and moisture) and weather (air temperature, rainfall, solar radiation, among others) were submitted to descriptive and multivariate statistical analysis in order to better understand the sugarcane ripening process and the relationship between this process and environmental conditions. The meteorological variables of best fit were used for elaborating the models to describe the ripening process, which were grouped into early, middle and late, according to the maturity pattern of the studied cultivars. The evaluation of the models was based on the analysis of residuals (presence of outliers, homogeneity of variances and normality of residuals) and on the comparison between estimated and independent data (mean errors, R2, agreement index of Willmott and confidence index of Camargo). Regardless of the pattern of sugarcane maturation (early, middle and late), the models that presented the best performance were for ATR, Brix, Pol% cane and humidity variables, with adjusted R2 values above 0.9 (p significant at 1%) and excellent performance, with C index above 0.85, when evaluated with independent data. The fiber model had the lowest adjusted R2 values (around 0.65) and problems in the distribution of its residuals (non-normality); however, when it was tested with independent data an acceptable performance was found, with C index above 0.8. The generated models showed that precipitation can be used as the main variable for predicting sugarcane quality and that its use can be extended to other areas than the one where it was originally generated.

Coagrupamento é uma estratégia para análise de dados capaz de encontrar grupos de dados, então denominados cogrupos, que são formados considerando subconjuntos diferentes das características descritivas dos dados. Contextos de aplicação caracterizados por apresentar subjetividade, como mineração de texto, são candidatos a serem submetidos à estratégia de coagrupamento; a flexibilidade em associar textos de acordo com características parciais representa um tratamento adequado a tal subjetividade. Um método para implementação de coagrupamento capaz de lidar com esse tipo de dados é a fatoração de matrizes. Nesta dissertação de mestrado são propostas duas estratégias para coagrupamento baseadas em fatoração de matrizes não-negativas, capazes de encontrar cogrupos organizados com sobreposição de colunas em uma matriz de valores reais positivos. As estratégias são apresentadas em termos de suas definições formais e seus algoritmos para implementação. Resultados experimentais quantitativos e qualitativos são fornecidos a partir de problemas baseados em conjuntos de dados sintéticos e em conjuntos de dados reais, sendo esses últimos contextualizados na área de mineração de texto. Os resultados são analisados em termos de quantização do espaço e capacidade de reconstrução, capacidade de agrupamento utilizando as métricas índice de Rand e informação mútua normalizada e geração de informação (interpretabilidade dos modelos). Os resultados confirmam a hipótese de que as estratégias propostas são capazes de descobrir cogrupos com sobreposição de forma natural, e que tal organização de cogrupos fornece informação detalhada, e portanto de valor diferenciado, para as áreas de análise de agrupamento e mineração de texto / Coclustering is a data analysis strategy which is able to discover data clusters, known as coclusters. This technique allows data to be clustered based on different subsets defined by data descriptive features. Application contexts characterized by subjectivity, such as text mining, are candidates for applying coclustering strategy due to the flexibility to associate documents according to partial features. The coclustering method can be implemented by means of matrix factorization, which is suitable to handle this type of data. In this thesis two strategies are proposed in non-negative matrix factorization for coclustering. These strategies are able to find column overlapping coclusters in a given dataset of positive data and are presented in terms of their formal definitions as well as their algorithms\' implementation. Quantitative and qualitative experimental results are presented through applying synthetic datasets and real datasets contextualized in text mining. This is accomplished by analyzing them in terms of space quantization, clustering capabilities and generated information (interpretability of models). The well known external metrics Rand index and normalized mutual information are used to achieve the analysis of clustering capabilities. Results confirm the hypothesis that the proposed strategies are able to discover overlapping coclusters naturally. Moreover, these coclusters produced by the new algorithms provide detailed information and are thus valuable for future research in cluster analysis and text mining