An unsupervised approach to feature discretization and selection

Many learning problems require handling high dimensional datasets with a relatively small number of instances. Learning algorithms are thus confronted with the curse of dimensionality, and need to address it in order to be effective. Examples of these types of data include the bag-of-words representation in text classification problems and gene expression data for tumor detection/classification. Usually, among the high number of features characterizing the instances, many may be irrelevant (or even detrimental) for the learning tasks. It is thus clear that there is a need for adequate techniques for feature representation, reduction, and selection, to improve both the classification accuracy and the memory requirements. In this paper, we propose combined unsupervised feature discretization and feature selection techniques, suitable for medium and high-dimensional datasets. The experimental results on several standard datasets, with both sparse and dense features, show the efficiency of the proposed techniques as well as improvements over previous related techniques.

A cultura, os valores contabilísticos e o julgamento profissional dos preparadores e dos revisores/auditores portugueses

Mestrado em Auditoria

Feature Discretization with Relevance and Mutual Information Criteria

Feature discretization (FD) techniques often yield adequate and compact representations of the data, suitable for machine learning and pattern recognition problems. These representations usually decrease the training time, yielding higher classification accuracy while allowing for humans to better understand and visualize the data, as compared to the use of the original features. This paper proposes two new FD techniques. The first one is based on the well-known Linde-Buzo-Gray quantization algorithm, coupled with a relevance criterion, being able perform unsupervised, supervised, or semi-supervised discretization. The second technique works in supervised mode, being based on the maximization of the mutual information between each discrete feature and the class label. Our experimental results on standard benchmark datasets show that these techniques scale up to high-dimensional data, attaining in many cases better accuracy than existing unsupervised and supervised FD approaches, while using fewer discretization intervals.

Comparação de métodos de identificação de bactérias floculentas presentes em ETAR

Dissertação apresentada na Faculdade de Ciências e Tecnologia da Universidade Nova de Lisboa para obtenção do grau de Mestre em Engenharia do Ambiente – Perfil Engenharia Sanitária

Using data mining techniques to support breast cancer diagnosis

More than ever, there is an increase of the number of decision support methods and computer aided diagnostic systems applied to various areas of medicine. In breast cancer research, many works have been done in order to reduce false-positives when used as a double reading method. In this study, we aimed to present a set of data mining techniques that were applied to approach a decision support system in the area of breast cancer diagnosis. This method is geared to assist clinical practice in identifying mammographic findings such as microcalcifications, masses and even normal tissues, in order to avoid misdiagnosis. In this work a reliable database was used, with 410 images from about 115 patients, containing previous reviews performed by radiologists as microcalcifications, masses and also normal tissue findings. Throughout this work, two feature extraction techniques were used: the gray level co-occurrence matrix and the gray level run length matrix. For classification purposes, we considered various scenarios according to different distinct patterns of injuries and several classifiers in order to distinguish the best performance in each case described. The many classifiers used were Naïve Bayes, Support Vector Machines, k-nearest Neighbors and Decision Trees (J48 and Random Forests). The results in distinguishing mammographic findings revealed great percentages of PPV and very good accuracy values. Furthermore, it also presented other related results of classification of breast density and BI-RADS® scale. The best predictive method found for all tested groups was the Random Forest classifier, and the best performance has been achieved through the distinction of microcalcifications. The conclusions based on the several tested scenarios represent a new perspective in breast cancer diagnosis using data mining techniques.

Descending pain modulation in a Kaolin-induced hydrocephalus rat model

Pain transmission at the spinal cord is modulated by descending actions that arise from supraspinal areas which collectively form the endogenous pain control system. Two key areas involved of the endogenous pain control system have a circunventricular location, namely the periaqueductal grey (PAG) and the locus coeruleus (LC). The PAG plays a crucial role in descending pain modulation as it conveys the input from higher brain centers to the spinal cord. As to the LC, it is involved in descending pain inhibition by direct noradrenergic projections to the spinal cord. In the context of neurological defects, several diseases may affect the structure and function of the brain. Hydrocephalus is a congenital or acquired disease characterized by an enlargement of the ventricles which leads to a distortion of the adjacent tissues, including the PAG and LC. Usually, patients suffering from hydrocephalus present dysfunctions in learning and memory and also motor deficits. It remains to be evaluated if lesions of the periventricular brain areas involved in pain control during hydrocephalus may affect descending pain control and, herein, affect pain responses. The studies included in the present thesis used an experimental model of hydrocephalus (the rat injected in the cisterna magna with kaolin) to study descending modulation of pain, focusing on the two circumventricular regions referred above (the PAG and the LC). In order to evaluate the effects of kaolin injection into the cisterna magna, we measured the degree of ventricular dilatation in sections encompassing the PAG by standard cytoarquitectonic stanings (thionin staining). For the LC, immunodetection of the noradrenaline-synthetizing enzyme tyrosine hydroxylase (TH) was performed, due to the noradrenergic nature of the LC neurons. In general, rats with kaolin-induced hydrocephalus presented a higher dilatation of the 4th ventricle, along with a tendency to a higher area of the PAG. Due to the validated role of detection the c-fos protooncogene as a marker of neuronal activation, we also studied neuronal activation in the several subnuclei which compose the PAG, namely the dorsomedial, dorsolateral, lateral and ventrolateral (VLPAG) parts. A decrease in the numbers of neurons immunoreactive for Fos protein (the product of activation of the c-fos protooncogene) was detected in rats injected with kaolin, whereas the remaining PAG subnuclei did not present changes in Fos-immunoreactive nuclei. Increases in the levels of TH in the LC, namely at the rostral parts of the nucleus, were detected in hydrocephalic animals. The following pain-related parameters were measured, namely 1) pain behavioural responses in a validated pain inflammatory test (the formalin test) and 2) the nociceptive activation of spinal cord neurons. A decrease in behavioral responses was detected in rats with kaolin-induced hydrocephalus was detected, namely in the second phase of the test (inflammatory phase). This is the phase of the formalin test in which the motor behaviour is less important, which is important since a semi-quantitative analysis of the motor performance of rats injected with kaolin indicates that these animals may present some motor impairments. Collectively, the results of the behavioral studies indicate that rats with kaolin-induced hydrocephalus exhibit hypoalgesia. A decrease in Fos expression was detected at the superficial dorsal layers of the spinal cord in rats with kaolin-induced hydrocephalus, further indicating that hydrocephalus decreases nociceptive responses. It remains to be ascertained if this is due to alterations in the PAG and LC in the rats with kaolin-induced hydrocephalus, which may affect descending pain modulation. It remains to be evaluated what are the mechanisms underlying the increased pain inhibition at the spinal dorsal horn in the hydrocephalus rats. Regarding the VLPAG, the decrease in neuronal activity may impair descending modulation. Since the LC has higher levels of TH in rats with kaolininduced hydrocephalus, which also appears to increase the noradrenergic innervation in the spinal dorsal horn, it is possible that an increase in the release of noradrenaline at the spinal cord accounts for pain inhibition. Our studies also determine the need to study in detail patients with hydrocephalus namely in what concerns their thresholds to pain and to perform imaging studies focused on the structure and function of pain control areas in the brain.

Evaluation of the Chagas Disease Control Program in Açucena Municipality, Rio Doce Valley, State of Minas Gerais, Brazil

Introduction Açucena Municipality, Rio Doce Valley, State of Minas Gerais, Brazil temporarily (2001-2005) interrupted epidemiological surveillance for Chagas disease. The objective of this work was to evaluate the Chagas Disease Control Program (CDCP) in Açucena and to offer suggestions for improving local epidemiological surveillance. Methods This study was conducted in three phases: I) a serological investigation of schoolchildren aged 5 to 15 years using an enzyme-linked immunosorbent assay (ELISA) test performed on blood collected on filter paper followed by ELISA, indirect immunofluorescence (IIF) and indirect hemaglutination (IHA) on venous blood for borderline cases and those in the gray zone of reactivity; II) vector evaluation using the data obtained by local health agents during 2006-2010; and III) examination by ELISA, IIF and IHA of serum samples from the inhabitants of houses where infected Triatoma vitticeps was found and evaluation of their knowledge about Chagas disease. Results Five individuals had inconclusive results in the ELISA screening but were seronegative for Chagas disease. The triatomine evaluation revealed the presence of three species: Triatoma vitticeps, Panstrongylus megistus and Panstrongylus diasi. Triatoma vitticeps was the most prevalent and widespread, with a higher (67%) index of Trypanosoma cruzi flagellates and evidence of colonization. Most of the inhabitants of the infested houses recognized triatomines and had basic knowledge about Chagas disease. Conclusions Although T. vitticeps is not clearly associated with Chagas disease transmission, these results highlight the importance of maintaining CDCP in endemic areas and the need for greater emphasis on epidemiological surveillance, especially in areas with important vectorial changes or that have been modified by human intervention.

Does the existing Corporate Governance structure and strict Legal Framework, affect the strategic decisionmaking of CEOs?

This thesis is a case study on Corporate Governance and Business Ethics, using the Portuguese Corporate Law as a general setting. The thesis was conducted in Portugal with illustrations on past cases under the Business Judgment Rule of the State of Delaware, U.SA along with illustrations on current cases in Portugal under the Portuguese Judicial setting, along with a comparative analysis between both. A debate is being considered among scholars and executives; a debate on best practices within corporate governance and corporate law, associated with recent discoveries of unlawful investments that lead to the bankruptcy of leading institutions and an aggravation of the crisis in Portugal. The study aimed at learning possible reasons and causes for the current situation of the country’s corporations along with attempts to discover the best way to move forward. From the interviews and analysis conducted, this paper concluded that the corporate governance structure and legal frameworks in Portugal were not the sole influencers behind the actions and decisions of Corporate Executives, nor were they the main triggers for the recent corporate mishaps. But it is rather a combination of different factors that played a significant role, such as cultural and ethical aspects, individual personalities, and others all of which created gray areas beyond the legal structure, which in turn accelerated and aggravated the corporate governance crisis in the country.

Physiographic and floristic gradients across topography in transitional seasonally dry evergreen forests of southeast Pará, Brazil

Seasonally dry evergreen forests in southeast Pará, Brazil are transitional between taller closed forests of the interior Amazon Basin and woodland savannas (cerrados) of Brazil's south-central plains. We describe abiotic and biotic gradients in this region near the frontier town of Redenção where forest structure and composition grade subtly across barely undulating topography. Annual precipitation averaged 1859 mm between 1995-2001, with nearly zero rainfall during the dry season months of June August. Annual vertical migrations of deep-soil water caused by seasonal rainfall underlie edaphic and floristic differences between high- and low-ground terrain. Low-ground soils are hydromorphic, shaped by perching water tables during the wet season, pale gray, brown, or white in color, with coarse texture, low moisture retention during the dry season, and relatively high macro-nutrient status in the surface horizons. Forest canopies on low ground are highly irregular, especially along seasonal streams, while overstory community composition differs demonstrably from that on high ground. High-ground soils are dystrophic, well-drained through the wet season, brown or red-yellow in color, with finer texture, higher moisture retention, and low macro-nutrient status in the surface horizons compared to low-ground soils. Forest canopies are, on average, taller, more regular, and more closed on high ground. Low-ground areas can be envisioned as energy and nutrient sinks, where, because of hydrologic cycles, canopy disturbance likely occurs more frequently than at high-ground positions if not necessarily at larger scales.

Avaliação de indicadores de qualidade do solo sob alternativas de recuperação do solo no Nordeste Paraense

Os indicadores convencionais de fertilidade de solo, mais comumente utilizados, enfatizam o balanço ácido-base e o conteúdo de elementos químicos, mas não consideram adequadamente as propriedades biológicas do solo, responsáveis por grande parte do seu funcionamento. Os indicadores biológicos, ainda pouco difundidos, levam em consideração a atividade microbiana do solo que é mais sensível à dinâmica do solo. Em Marapanim-PA, e utilizando um desenho experimental inteiramente casualizado, com quatro repetições, avaliou-se o comportamento dos indicadores do solo carbono e nitrogênio da biomassa microbiana, fosfatase ácida, matéria orgânica, carbono orgânico e fósforo disponível, em um solo com cultivo sequencial de mandioca e milho, preparo com corte-e-trituração da capoeira original. Os fatores corte-e-trituração, fosfato natural e adubação verde foram arranjados nos seguintes tratamentos: capoeira; testemunha; fosfato natural (FN); feijão-de-porco [Canavalia ensiformis (L.) DC] (FP); FN + FP; FN + guandu [Cajanus cajan (L.) Millspaugh]. (G); FN + titônia [Tithonia diversifolia (Hemsley) A. Gray] (T); FP + G; FP + T. Em quatro épocas de amostragens, foram tiradas amostras compostas do solo (0 - 10 cm) de cada parcela para análise dos indicadores. Os indicadores estudados foram influenciados pela época de amostragem do solo. Os indicadores fosfatase ácida, nitrogênio da biomassa microbiana e carbono da biomassa microbiana foram capazes de detectar os efeitos de maior número de tratamentos de recuperação do solo testados, enquanto a matéria orgânica do solo foi o menos capaz. Ao longo do tempo, a resposta dos indicadores variou em função dos fatores estudados.

Caracterização de uso e cobertura da terra na Amazônia utilizando imagens duais multitemporais do COSMO-SkyMed

A utilização de imagens de radar é fonte alternativa de informações para subsidiar o monitoramento da região amazônica, visto que as imagens ópticas têm limitações de imageamento em zonas tropicais face a ocorrência de nuvens. Por conseguinte este trabalho teve como objetivo analisar a capacidade das imagens-radar de banda X multitemporais e polarizadas obtidas pelo satélite COSMO-SkyMed (COnstellation of small Satellites for Mediterranean basin Observation), no modo intensidade, isoladamente e agregados às informações texturais, na caracterização temática de uso e cobertura da terra no município de Humaitá/AM. A metodologia empregada consistiu da: análise das imagens duais obtidas em duas aquisições subsequentes, de forma a explorar a potencialidade do conjunto de dados na forma quad-pol intensidade; extração dos atributos texturais a partir da matriz de coocorrência (Gray Level Co-occurrence Matrix) e posterior classificação contextual; avaliação estatística de desempenho temático das imagens intensidade e texturais, isoladas e em grupos polarizados. Dentre os vários resultados alcançados, foi verificado que o grupo formado somente pelas imagens intensidade apresentou o melhor desempenho, comparado àqueles contendo os atributos texturais. Nesta separabilidade, estavam envolvidas as classes de floresta, floresta aluvial, reflorestamento, savana, pasto e queimada, obtendo-se 66% de acurácia total e valor Kappa de 0,55. Os resultados mostraram que as imagens de banda X do COSMO-SkyMed, modo StripMap (Ping-Pong), multipolarizadas, têm potencial moderado para a caracterização e monitoramento da dinâmica de uso e cobertura da terra na Amazônia brasileira.

A bio-inspired computational model for motion detection

Tese de Doutoramento (Programa Doutoral em Engenharia Biomédica)

Investigations on relational control of time perception

Tese de Doutoramento em Psicologia Básica

Characterization of late antique and early medieval pottery production of the city of Braga and its territory

The fundamental goal we set ourselves when developing this study is to try to characterize, both technically and formally, ceramics made in the city of Braga and its territory from the initial moments of the Late Antiquity to the Middle Ages. Thus, we will focus on analyzing some own productions that appear attached to the phases of late antique occupation —ceramics of red engobes and late gray—, as well as in the early medieval containers identified in different archaeological interventions practiced in the Braga environment. Concretely, we will analyze the material from various excavations conducted recently at the Theatre in the solar number 20/28 and 36/56 from the Afonso Henriques Street and the former District Hostel as well as the church of São Martinho de Dume.