Rozprawy doktorskie na temat „Rage Against The Machine”

This dissertation addresses the phenomenon of 'radical commodities'-commercial products which advance an oppositional politics. Examples of such include the products of Rage Against The Machine, a 'revolutionary' rock band; Michael Moore, a best-selling author and award-winning documentary filmmaker; Naomi Klein, a journalist and author of the international bestseller No Logo; The Body Shop, a multinational manufacturer and retailer of 'natural' cosmetics and toiletries; Freitag, a company which manufactures bags, wallets and other fashionable accessories from recycled materials, and; the Adbusters Media Foundation, publisher of Adbusters magazine and producer of Blackspot shoes. Radical commodities are fundamentally paradoxical objects whose apparent ethic would appear to be at odds with the fact that they are commodities. This dissertation asks: can a commodity-object legitimately serve as a vehicle for social and political critique? It is reasoned that the problem of radical commodities is principally structural. Marx's seminal writings on the commodity accordingly represent the logical point of departure. The Marxian analysis illuminates not only the commodity-structure, but also the political problematic which emerges from that structure-for Marx, the commodity is a mechanism of exploitation. From an orthodox Marxist perspective, the idea of a radical commodity would therefore be most contradictory, or indeed impossible. It is argued, however, that the Marxian analysis is inconclusive. This dissertation traces a genealogy of analyses of the commodity, which variously advance or diverge from the orthodox Marxist position. From a perspective of the consumption of commodity-objects, the radical commodity would appear to be possible. Yet, the relationship between the commodity-structure and the capitalist ideology runs deep. The question of the radical commodity is therefore markedly more complex than it might initially appear. With regard to the ideological consequence of the commodity-structure, however, certain streams of post-Marxist analysis are themselves problematic, for they ultimately short-circuit historical critique and destabilise the very possibility of politics. In contrast, this dissertation seeks to reaffirm a place for politics and, in so doing, to establish the theoretical possibility of radical commodities. To contend that the idea of a radical commodity is not fundamentally contradictory, however, says nothing of the political potency of such objects. These are undoubtedly complex objects, whose peculiarities cannot be ascertained by abstract theorisation alone. For this reason, this dissertation also employs empirical analyses of a number of radical commodities. In sum, it is argued that the sphere of commodities should be admitted as a possible site for the expression or implementation of a radical politics, and thus that radical commodities should be understood as a legitimate vehicle for social and political critique, but that such objects are by no means free from contradiction, and that the political efficacy of these products is anything but guaranteed.

Denna studie undersöker hur måleriskt arbete fokuserat på kvalitéer utöver avbildande, funktion och skönhet kan tillgängliggöras och förstärkas inom det digitala måleriprogrammet Photoshop. En pedagogisk metod och den digitala miljön (”workspacet”) analyseras och optimeras för syftet. Diskurser inom digital bildkonst analyseras och synliggörs för att ytterligare möjliggöra en konstnärlig frihet på fältet. Arbetet syftar till att vara en grund för bildpedagogisk undervisning och uppgiftsbyggande. Studiens empiri genererades av ett småskaligt cykliskt måleriexperiment utfört i Photoshop. Studien finner att måleriskt arbete i Photoshop överlag kan tillgängliggöras och förstärkas genom att låta problemlösande och innoverande driva kunskapandet, samt ger ett förslag på hur ett specialanpassat måleriworkspace kan vara uppbyggt. Vidare finner den även att det är möjligt att öka tillgängligheten till och förstärka fokus på kvalitéer utöver avbildande, funktion och skönhet inom digitalt bildskapande genom att använda visuell diskursanalys som motståndspraktik.

This master’s thesis offers a descriptive investigation of how public Euroscepticism has changed during the economic crisis. By regarding Eurosceptic opinions as positions on a scale as well as differentiating between opposition towards policy (concrete decisions and measures) and polity (the EU as a political system), the results show that overall, European citizens have not only become negative towards how the EU handled the crisis in terms of policy output. Policy opposition also tends to spill over into the polity area, entailing increasing preferences for reducing EU competences and for leaving the union all together. The policy-polity distinction is a perspective on Euroscepticism as opposition, that has largely been set aside in previous research. Furthermore, the paper describes how support for the EU polity and its policies has changed in different member states through a cross country analysis. The results show that there is great and unexpected variation between countries that would be assumed to display a similar development in the light of previous research. In addition, there are similarities between countries that were expected to develop differently. This indicates a limited understanding in this research field of how national factors influence public attitudes towards the European Union. A topic that needs to be investigated further.

Traditional approaches in network intrusion detection follow a signature-based ap- proach, however the use of anomaly detection approaches based on machine learning techniques have been studied heavily for the past twenty years. The continuous change in the way attacks are appearing, the volume of attacks, as well as the improvements in the big data analytics space, make machine learning approaches more alluring than ever. The intention of this thesis is to show that using machine learning in the intrusion detection domain should be accompanied with an evaluation of its robustness against adversaries. Several adversarial techniques have emerged lately from the deep learning research, largely in the area of image classification. These techniques are based on the idea of introducing small changes in the original input data in order to make a machine learning model to misclassify it. This thesis follows a big data Analytics methodol- ogy and explores adversarial machine learning techniques that have emerged from the deep learning domain, against machine learning classifiers used for network intrusion detection. The study looks at several well known classifiers and studies their performance under attack over several metrics, such as accuracy, F1-score and receiver operating character- istic. The approach used assumes no knowledge of the original classifier and examines both general and targeted misclassification. The results show that using relatively sim- ple methods for generating adversarial samples it is possible to lower the detection accuracy of intrusion detection classifiers from 5% to 28%. Performance degradation is achieved using a methodology that is simpler than previous approaches and it re- quires only 6.25% change between the original and the adversarial sample, making it a candidate for a practical adversarial approach.

In the recent years, connected surveillance systems have been witnessing an unprecedented evolution owing to the advancements in internet of things and deep learning technologies. However, vulnerabilities to various kinds of attacks both at the cyber network-level and at the physical worldlevel are also rising. This poses danger not only to the devices but also to human life and property. The goal of this thesis is to enhance the security of an internet of things, focusing on connected video-based surveillance systems, by proposing multiple novel solutions to address security issues at the cyber network-level and to defend such systems at the physical world-level. In order to enhance security at the cyber network-level, this thesis designs and develops solutions to detect network intrusions in an internet of things such as surveillance cameras. The first solution is a novel method for network flow features transformation, named TempoCode. It introduces a temporal codebook-based encoding of flow features based on capturing the key patterns of benign traffic in a learnt temporal codebook. The second solution takes an unsupervised learning-based approach and proposes four methods to build efficient and adaptive ensembles of neural networks-based autoencoders for intrusion detection in internet of things such as surveillance cameras. To address the physical world-level attacks, this thesis studies, for the first time to the best of our knowledge, adversarial patches-based attacks against a convolutional neural network (CNN)- based surveillance system designed for vehicle make and model recognition (VMMR). The connected video-based surveillance systems that are based on deep learning models such as CNNs are highly vulnerable to adversarial machine learning-based attacks that could trick and fool the surveillance systems. In addition, this thesis proposes and evaluates a lightweight defense solution called SIHFR to mitigate the impact of such adversarial-patches on CNN-based VMMR systems, leveraging the symmetry in vehicles’ face images. The experimental evaluations on recent realistic intrusion detection datasets prove the effectiveness of the developed solutions, in comparison to state-of-the-art, in detecting intrusions of various types and for different devices. Moreover, using a real-world surveillance dataset, we demonstrate the effectiveness of the SIHFR defense method which does not require re-training of the target VMMR model and adds only a minimal overhead. The solutions designed and developed in this thesis shall pave the way forward for future studies to develop efficient intrusion detection systems and adversarial attacks mitigation methods for connected surveillance systems such as VMMR.

Master of Science
Department of Biological & Agricultural Engineering
Naiqian Zhang
The primary objective of this research is to achieve automatic evaluation of the efficiency of using Trichogramma bourarachae for biological control of Cadra (=Ephestia) cautella by calculating the rate of parasitization. Cadra cautella is a moth feeding as a larva on dried fruit as well as stored nuts, seeds, and other warehouse foodstuffs. It attacks dates from ripening stages while on tree, throughout storage, and until consumption. These attacks cause significant qualitative and quantitative damages, which negatively affect dates’ marketability, resulting in economic losses. To achieve this research goal, tasks were accomplished by developing image processing algorithms for detecting, identifying, and differentiating between three Cadra cautella egg categories based on the success of Trichogramma parasitization against them. The egg categories were parasitized (black and dark red), fertile (unhatched yellow), and hatched (white) eggs. Color, intensity, and shape information was obtained from digital images of Cadra eggs after they were subjected to Trichogramma parasitization and used to develop detection algorithms. Two image processing methods were developed. The first method included segmentation and extractions of color and morphological features followed by watershed delineation, and is referred to as the "Watershed Method" (WT). The second method utilized the Hough Transformation to find circular objects followed by convolution filtering, and is referred to as the "Hough Transform Method" (HT). The algorithms were developed based on 2 images and then tested on more than 40 images. The WT and the HT methods achieved correct classification rates (CCRs) of parasitized eggs of 92% and 96%, respectively. Their CCRs of yellow eggs were 48% and 94%, respectively, while for white eggs the CCRs were 42% and 73%. Both methods performed satisfactorily in detecting the parasitized eggs, but the HT outperformed the WT in detecting the unparasitized eggs. The developed detection methods will enable automatic evaluation of biological control of Cadra (=Ephestia) cautella using Trichogramma bourarachae. Moreover, with few adjustments these methods can be used in similar applications such as detecting plant diseases in terms of presence of insects or their eggs.

Natural killer (NK) cells are lymphocytes that comprise part of the innate immune system and play a key role in the early defence against pathogenic organisms and cancer. CpG oligodeoxynucelotides (ODNs) are short synthetic ODN containing unmethylated CpG dinucleotide motifs that have immune-enhancing effects. NK cell-derived IFN-γ is essential for the effects of CpG ODNs, but how NK cells become activated by CpG ODNs remains unclear. We found that CpG ODN-mediated stimulation of NK cells requires IL-12 or IL-18. CpG ODNs did not stimulate IL-12-deficient mouse spleen cells and IL-12 neutralization almost completely inhibited IFN-γ production. Although IL-18 was undetectable in cultures, neutralization significantly dampened the IFN-γ response and addition of exogenous IL-18 greatly enhanced CpG ODN-mediated NK cell stimulation. IL-12 is mainly produced by Gr-1⁺ monocytes and neutrophils, while what cells produce IL-18 remains unknown. We then tested the anti-leukemia effects of CpG ODN-stimulated NK cells. Studies with human acute myeloid leukemia (AML) patients have shown that haploidentical NK cells effectively kill AML blasts, but their ability to lyse leukemia initiating cells (LICs) has not been studied. Therefore, we tested NK cells from haploidentical F1 mice against the mouse AML cell line MN1. F1 mouse NK cells expanded in cultures in the presence of IL-15 and stimulated by CpG ODNs plus IL-18, effectively killed bulk MN1 cells in vitro and reduced the numbers of in vitro colony forming cells. NK cell-treated MN1 cells were also injected into irradiated B6 mice to test whether AML LICs were also killed. F1 mouse NK cells seemed to kill some AML initiating cells since mice receiving NK-treated MN1 cells survived significantly longer than those given untreated MN1 cells, but the frequency of LICs did not significantly differ between MN1 cells incubated with or without NK cells. For NK cells to be used as a treatment for AML, we must find a way to induce a higher cytotoxicity in NK cells or to target them specifically towards LIC.

Virtualisation is a technology that is more and more applied due to its advantages regarding cost and operation. It is often believed that it provides a better security for an IT environment since it enables centralisation of hardware. However, virtualisation changes an IT environment fundamentally and contains new vulnerabilities that must be considered. It is of interest to evaluate whether the belief that virtual environments provide a better security for an IT environment is true or not. In this project, the resiliency against attacks for physical environments and virtual environments is analysed to determine which one provides a higher resiliency and why. Therefore, the physical and digital attack surfaces of all entities are analysed to reveal the relevant vulnerabilities that could be exploited. Beside a theoretical research, a physical and a virtual environment have been established to test chosen attacks practically. The results show that virtual environments are less resilient than physical environments, especially to common attacks. This shows that virtualisation is still a technology that is new to many companies and the vulnerabilities it has must be taken seriously.

Context. Machine learning is a constantly developing subfield within the artificial intelligence field. The number of domains in which we deploy machine learning models is constantly growing and the systems using these models spread almost unnoticeably in our daily lives through different devices. In previous years, lots of time and effort has been put into increasing the performance of these models, overshadowing the significant risks of attacks targeting the very core of the systems, the trained machine learning models themselves. A specific attack with the aim of fooling the decision-making of a model, called the adversarial input attack, has almost exclusively been researched for models processing image data. However, the threat of adversarial input attacks stretches beyond systems using image data, to e.g the tabular domain which is the most common data domain used in the industry. Methods used for interpreting complex machine learning models can help humans understand the behavior and predictions of these complex machine learning systems. Understanding the behavior of a model is an important component in detecting, understanding and mitigating vulnerabilities of the model. Objectives. This study aims to reduce the research gap of adversarial input attacks and defenses targeting machine learning models in the tabular data domain. The goal of this study is to analyze how model-agnostic interpretation methods can be used in order to mitigate and detect adversarial input attacks on tabular data. Methods. The goal is reached by conducting three consecutive experiments where model interpretation methods are analyzed and adversarial input attacks are evaluated as well as visualized in terms of perceptibility. Additionally, a novel method for adversarial input attack detection based on model interpretation is proposed together with a novel way of defensively using feature selection to reduce the attack vector size. Results. The adversarial input attack detection showed state-of-the-art results with an accuracy over 86%. The proposed feature selection-based mitigation technique was successful in hardening the model from adversarial input attacks by reducing their scores by 33% without decreasing the performance of the model. Conclusions. This study contributes with satisfactory and useful methods for adversarial input attack detection and mitigation as well as methods for evaluating and visualizing the imperceptibility of attacks on tabular data.
Kontext. Maskininlärning är ett område inom artificiell intelligens som är under konstant utveckling. Mängden domäner som vi sprider maskininlärningsmodeller i växer sig allt större och systemen sprider sig obemärkt nära inpå våra dagliga liv genom olika elektroniska enheter. Genom åren har mycket tid och arbete lagts på att öka dessa modellers prestanda vilket har överskuggat risken för sårbarheter i systemens kärna, den tränade modellen. En relativt ny attack, kallad "adversarial input attack", med målet att lura modellen till felaktiga beslutstaganden har nästan uteslutande forskats på inom bildigenkänning. Men, hotet som adversarial input-attacker utgör sträcker sig utom ramarna för bilddata till andra datadomäner som den tabulära domänen vilken är den vanligaste datadomänen inom industrin. Metoder för att tolka komplexa maskininlärningsmodeller kan hjälpa människor att förstå beteendet hos dessa komplexa maskininlärningssystem samt de beslut som de tar. Att förstå en modells beteende är en viktig komponent för att upptäcka, förstå och mitigera sårbarheter hos modellen. Syfte. Den här studien försöker reducera det forskningsgap som adversarial input-attacker och motsvarande försvarsmetoder i den tabulära domänen utgör. Målet med denna studie är att analysera hur modelloberoende tolkningsmetoder kan användas för att mitigera och detektera adversarial input-attacker mot tabulär data. Metod. Det uppsatta målet nås genom tre på varandra följande experiment där modelltolkningsmetoder analyseras, adversarial input-attacker utvärderas och visualiseras samt där en ny metod baserad på modelltolkning föreslås för detektion av adversarial input-attacker tillsammans med en ny mitigeringsteknik där feature selection används defensivt för att minska attackvektorns storlek. Resultat. Den föreslagna metoden för detektering av adversarial input-attacker visar state-of-the-art-resultat med över 86% träffsäkerhet. Den föreslagna mitigeringstekniken visades framgångsrik i att härda modellen mot adversarial input attacker genom att minska deras attackstyrka med 33% utan att degradera modellens klassifieringsprestanda. Slutsats. Denna studie bidrar med användbara metoder för detektering och mitigering av adversarial input-attacker såväl som metoder för att utvärdera och visualisera svårt förnimbara attacker mot tabulär data.

Are robots stealing jobs? A large majority of Europeans think they do. This paper shedslight on the drivers of anxiety of the European population about automation and possiblejob replacement. We find that fears of automation are linked to individual characteristics –“who you are” – but also macro ones – “where you’re born”. Our results show that women,less educated individuals, blue-collar workers and those that are not skilled in ICT are moreconcerned with robotization, but younger individuals fear for their own jobs the most. Atthe country-level, economies with higher shares of employment in manufacturing industryand agriculture, higher GDP growth and higher unemployment rates are more likely to fearthe impact of new technologies, hinting at a role of both structural and cyclical factors.

This study is a cultural investigation into rape, with specific focus on the role of African men. With more than 70 000 cases of rape and sexual violence reported in a year in South Africa, and estimations that this may reflect one ninth of the actual number only, South Africa has been labelled the “rape capital” of the world. The study seeks to explain the root causes of rape, its ontological make-up and possibilities for resolving the issue by identifying cultural aspects, factors and manifestations that either justify or sanction rape. Four concepts, namely, rape, masculinity, culture and rage, serve as the thematic lens for identifying and interrogating cultural conditions through multidisciplinary and Africancentred perspectives. The analyses contained in the study are based on a mapping process involving comparing the data from a wide range of literature and also focus group interviews. Highlighting the multi-layered complexities of rape as phenomena, the study then outlines recommendations for transformative work in research, cultural institutions, communities, families and men
Educational Studies
M. Ed. (Philosophy of Education)

This study is a cultural investigation into rape, with specific focus on the role of African men. With more than 70 000 cases of rape and sexual violence reported in a year in South Africa, and estimations that this may reflect one ninth of the actual number only, South Africa has been labelled the “rape capital” of the world. The study seeks to explain the root causes of rape, its ontological make-up and possibilities for resolving the issue by identifying cultural aspects, factors and manifestations that either justify or sanction rape. Four concepts, namely, rape, masculinity, culture and rage, serve as the thematic lens for identifying and interrogating cultural conditions through multidisciplinary and Africancentred perspectives. The analyses contained in the study are based on a mapping process involving comparing the data from a wide range of literature and also focus group interviews. Highlighting the multi-layered complexities of rape as phenomena, the study then outlines recommendations for transformative work in research, cultural institutions, communities, families and men
Educational Studies
M. Ed. (Philosophy of Education)

Gradient-based adversarial attacks on neural networks threaten extremely critical applications such as medical diagnosis and biometric authentication. These attacks use the gradient of the neural network to craft imperceptible perturbations to be added to the test data, in an attempt to decrease the accuracy of the network. We propose a defense to combat such attacks, which can be modified to reduce the training time of the network by as much as 71%, and can be further modified to reduce the training time of the defense by as much as 19%. Further, we address the threat of uncertain behavior on the part of the attacker, a threat previously overlooked in the literature that considers mostly white box scenarios. To combat uncertainty on the attacker's part, we train our defense with an ensemble of attacks, each generated with a different attack algorithm, and using gradients of distinct architecture types. Finally, we discuss how we can prevent the attacker from breaking the defense by estimating the gradient of the defense transformation.

碩士
國立清華大學
化學系
104
Human S100A9 (Calgranulin B) is a Ca2+-binding protein from the S100 family that often accompanies human the S100A8 protein (Calgranulin A). S100A9 presents as a homodimer in myeloid cells and emerges as an important mediator during inflammation after calcium binds to its EF hand motifs. Human receptors for the advanced glycation end products (RAGE) protein is one of the target proteins for S100A9 binding to its hydrophobic surface. Interaction between these two proteins will trigger signaling transduction cascades that promote cell growth, proliferation, and tumorigenesis. Here, we solved the solution structure of the mS100A9 homodimer by conducting multi-dimensional NMR experiments. We further characterize the solution interactions between mS100A9 and the RAGE V domain as well as S100A9 with the CHAPS molecule via NMR spectroscopy. Finally, using the HADDOCK program, we demonstrate that CHAPS molecules play a crucial role in blocking the interaction between mS100A9 and the RAGE V domain. WST-1 assay results also support that CHAPS inhibit the bioactivity of mS100A9. This report will help to inform new drug development against cell proliferation.

This study followed a qualitative research approach, and semi-structured interviews regarding the subject matter were conducted with the knowledgeable and experienced respondents in the Gauteng traffic-related departments. A literature review was also conducted to provide a comprehensive understanding of the research problem in both local and international context. The research stresses the fact that motorists are continuing to lose their lives on Gauteng province, as a result of violent traffic disputes, therefore people suffer financial, physical, psychological as well as social effects as a consequence of such actions. The findings revealed that all the traffic stakeholders are working cooperatively towards implementing the crime prevention strategic plans, but for policing road rage in particular there is no specific strategy in action. In addition, this study reveals that it is impossible for the traffic police to curb road rage incidents because the latter occur as a result of unpredictable human behavior.
Police Practice
M. Tech. (Policing)

碩士
國立屏東技術學院
資訊管理學系研究所
85
In recent years researchers have focused a significant amount of attention on scheduling problems in Flexible Manufacturing Systems (FMSs). Most traditional research in this area used only time-based rules and performance criteria. Ho wever, in most managerial problems, cost was a more global and homogeneous cri terion than time-based criteria such as flow-time and lateness. Therefore, min imizing total cost is a more appropriate objective.This article investigates t he scheduling problem of FMSs by using a simulation model of an FMS. Specifica lly, relative performance of scheduling rules for machines and automatedguided vehicles (AGVs) are analyzed against the total cost criterion. Furthermore, o ne new composite rules is proposed, using job due date, processing time, curre nt shop time, inventory holding cost, and tardiness cost information. The new rule is compared with five other well-known rules that are basically time-orie nted and cost-based. Results show that total cost dependent scheduling rules o utperform total cost independent rules.

Thesis (M.S.)--University of Wisconsin--Madison, 1997.
Typescript. eContent provider-neutral record in process. Description based on print version record. Includes bibliographical references (leaves 52-55).

Dissertation presented as the partial requirement for obtaining a Master's degree in Information Management, specialization in Knowledge Management and Business Intelligence
Over the past decades, financial institutions have been giving increased importance to credit risk management as a critical tool to control their profitability. More than ever, it became crucial for these institutions to be able to well discriminate between good and bad clients for only accepting the credit applications that are not likely to default. To calculate the probability of default of a particular client, most financial institutions have credit scoring models based on parametric techniques. Logistic regression is the current industry standard technique in credit scoring models, and it is one of the techniques under study in this dissertation. Although it is regarded as a robust and intuitive technique, it is still not free from several critics towards the model assumptions it takes that can compromise its predictions. This dissertation intends to evaluate the gains in performance resulting from using more modern non-parametric techniques instead of logistic regression, performing a model comparison over four different real-life credit datasets. Specifically, the techniques compared against logistic regression in this study consist of two single classifiers (decision tree and SVM with RBF kernel) and two ensemble methods (random forest and stacking with cross-validation). The literature review demonstrates that heterogeneous ensemble approaches have a weaker presence in credit scoring studies and, because of that, stacking with cross-validation was considered in this study. The results demonstrate that logistic regression outperforms the decision tree classifier, has similar performance in relation to SVM and slightly underperforms both ensemble approaches in similar extents.

abstract: The radar performance of detecting a target and estimating its parameters can deteriorate rapidly in the presence of high clutter. This is because radar measurements due to clutter returns can be falsely detected as if originating from the actual target. Various data association methods and multiple hypothesis filtering approaches have been considered to solve this problem. Such methods, however, can be computationally intensive for real time radar processing. This work proposes a new approach that is based on the unsupervised clustering of target and clutter detections before target tracking using particle filtering. In particular, Gaussian mixture modeling is first used to separate detections into two Gaussian distinct mixtures. Using eigenvector analysis, the eccentricity of the covariance matrices of the Gaussian mixtures are computed and compared to threshold values that are obtained a priori. The thresholding allows only target detections to be used for target tracking. Simulations demonstrate the performance of the new algorithm and compare it with using k-means for clustering instead of Gaussian mixture modeling.
Dissertation/Thesis
Masters Thesis Electrical Engineering 2016

With the global economic crisis, which reached its peak in the second half of 2008, and before a market shaken by economic instability, financial institutions have taken steps to protect the banks’ default risks, which had an impact directly in the form of analysis in credit institutions to individuals and to corporate entities. To mitigate the risk of banks in credit operations, most banks use a graded scale of customer risk, which determines the provision that banks must do according to the default risk levels in each credit transaction. The credit analysis involves the ability to make a credit decision inside a scenario of uncertainty and constant changes and incomplete transformations. This ability depends on the capacity to logically analyze situations, often complex and reach a clear conclusion, practical and practicable to implement. Credit Scoring models are used to predict the probability of a customer proposing to credit to become in default at any given time, based on his personal and financial information that may influence the ability of the client to pay the debt. This estimated probability, called the score, is an estimate of the risk of default of a customer in a given period. This increased concern has been in no small part caused by the weaknesses of existing risk management techniques that have been revealed by the recent financial crisis and the growing demand for consumer credit.The constant change affects several banking sections because it prevents the ability to investigate the data that is produced and stored in computers that are too often dependent on manual techniques. Among the many alternatives used in the world to balance this risk, the provision of guarantees stands out of guarantees in the formalization of credit agreements. In theory, the collateral does not ensure the credit return, as it is not computed as payment of the obligation within the project. There is also the fact that it will only be successful if triggered, which involves the legal area of the banking institution. The truth is, collateral is a mitigating element of credit risk. Collaterals are divided into two types, an individual guarantee (sponsor) and the asset guarantee (fiduciary). Both aim to increase security in credit operations, as an payment alternative to the holder of credit provided to the lender, if possible, unable to meet its obligations on time. For the creditor, it generates liquidity security from the receiving operation. The measurement of credit recoverability is a system that evaluates the efficiency of the collateral invested return mechanism. In an attempt to identify the sufficiency of collateral in credit operations, this thesis presents an assessment of smart classifiers that uses contextual information to assess whether collaterals provide for the recovery of credit granted in the decision-making process before the credit transaction become insolvent. The results observed when compared with other approaches in the literature and the comparative analysis of the most relevant artificial intelligence solutions, considering the classifiers that use guarantees as a parameter to calculate the risk contribute to the advance of the state of the art advance, increasing the commitment to the financial institutions.
Com a crise econômica global, que atingiu seu auge no segundo semestre de 2008, e diante de um mercado abalado pela instabilidade econômica, as instituições financeiras tomaram medidas para proteger os riscos de inadimplência dos bancos, medidas que impactavam diretamente na forma de análise nas instituições de crédito para pessoas físicas e jurídicas. Para mitigar o risco dos bancos nas operações de crédito, a maioria destas instituições utiliza uma escala graduada de risco do cliente, que determina a provisão que os bancos devem fazer de acordo com os níveis de risco padrão em cada transação de crédito. A análise de crédito envolve a capacidade de tomar uma decisão de crédito dentro de um cenário de incerteza e mudanças constantes e transformações incompletas. Essa aptidão depende da capacidade de analisar situações lógicas, geralmente complexas e de chegar a uma conclusão clara, prática e praticável de implementar. Os modelos de Credit Score são usados para prever a probabilidade de um cliente propor crédito e tornar-se inadimplente a qualquer momento, com base em suas informações pessoais e financeiras que podem influenciar a capacidade do cliente de pagar a dívida. Essa probabilidade estimada, denominada pontuação, é uma estimativa do risco de inadimplência de um cliente em um determinado período. A mudança constante afeta várias seções bancárias, pois impede a capacidade de investigar os dados que são produzidos e armazenados em computadores que frequentemente dependem de técnicas manuais. Entre as inúmeras alternativas utilizadas no mundo para equilibrar esse risco, destacase o aporte de garantias na formalização dos contratos de crédito. Em tese, a garantia não “garante” o retorno do crédito, já que não é computada como pagamento da obrigação dentro do projeto. Tem-se ainda, o fato de que esta só terá algum êxito se acionada, o que envolve a área jurídica da instituição bancária. A verdade é que, a garantia é um elemento mitigador do risco de crédito. As garantias são divididas em dois tipos, uma garantia individual (patrocinadora) e a garantia do ativo (fiduciário). Ambos visam aumentar a segurança nas operações de crédito, como uma alternativa de pagamento ao titular do crédito fornecido ao credor, se possível, não puder cumprir suas obrigações no prazo. Para o credor, gera segurança de liquidez a partir da operação de recebimento. A mensuração da recuperabilidade do crédito é uma sistemática que avalia a eficiência do mecanismo de retorno do capital investido em garantias. Para tentar identificar a suficiência das garantias nas operações de crédito, esta tese apresenta uma avaliação dos classificadores inteligentes que utiliza informações contextuais para avaliar se as garantias permitem prever a recuperação de crédito concedido no processo de tomada de decisão antes que a operação de crédito entre em default. Os resultados observados quando comparados com outras abordagens existentes na literatura e a análise comparativa das soluções de inteligência artificial mais relevantes, mostram que os classificadores que usam garantias como parâmetro para calcular o risco contribuem para o avanço do estado da arte, aumentando o comprometimento com as instituições financeiras.

Dissertation presented as the partial requirement for obtaining a Master's degree in Data Science and Advanced Analytics
Extensive recent research has shown the importance of innovation in medical healthcare, with a focus on Pneumonia. It is vital and lifesaving to predict Pneumonia cases as fast as possible and preferably in advance of the symptoms. An online database source managed to gather Pneumonia-specific image data, with not just the presence of the infection, but also the nature of it, divided in bacterial- and viral infection. The first achievement is extracting valuable information from the X-Ray image datasets. Using several ImageNet pre-trained CNNs, knowledge can be gained from images and transferred to numeric arrays. This, both binary and multi-class classification data, requires a sophisticated prediction algorithm that recognizes X-Ray image patterns. Multiple, recently performed experiments show promising results about the innovative Semantic Learning Machine (SLM) that is essentially a geometric semantic hill climber for feedforward Neural Networks. This SLM is based on a derivation of the Geometric Semantic Genetic Programming (GSGP) mutation operator for real-value semantics. To prove the outperformance of the binary and multi-class SLM in general, a selection of commonly used algorithms is necessary in this research. A comprehensive hyperparameter optimization is performed for commonly used algorithms for those kinds of real-life problems, such as: Random Forest, Support Vector Machine, KNearestNeighbors and Neural Networks. The results of the SLM are promising for the Pneumonia application but could be used for all types of predictions based on images in combination with the CNN feature extractions.
Uma extensa pesquisa recente mostrou a importância da inovação na assistência médica, com foco na pneumonia. É vital e salva-vidas prever os casos de pneumonia o mais rápido possível e, de preferência, antes dos sintomas. Uma fonte on-line conseguiu coletar dados de imagem específicos da pneumonia, identificando não apenas a presença da infecção, mas também seu tipo, bacteriana ou viral. A primeira conquista é extrair informações valiosas dos conjuntos de dados de imagem de raios-X. Usando várias CNNs pré-treinadas da ImageNet, é possível obter conhecimento das imagens e transferi-las para matrizes numéricas. Esses dados de classificação binários e multi-classe requerem um sofisticado algoritmo de predição que reconhece os padrões de imagem de raios-X. Vários experimentos realizados recentemente mostram resultados promissores sobre a inovadora Semantic Learning Machine (SLM), que é essencialmente um hill climber semântico geométrico para feedforward neural network. Esse SLM é baseado em uma derivação do operador de mutação da Geometric Semantic Genetic Programming (GSGP) para valor-reais semânticos. Para provar o desempenho superior do SLM binário e multi-classe em geral, é necessária uma seleção de algoritmos mais comuns na pesquisa. Uma otimização abrangente dos hiperparâmetros é realizada para algoritmos comumente utilizados para esses tipos de problemas na vida real, como Random Forest, Support Vector Machine,K-Nearest Neighbors and Neural Networks. Os resultados do SLM são promissores para o aplicativo pneumonia, mas podem ser usados para todos os tipos de previsões baseadas em imagens em combinação com as extrações de recursos da CNN.