Understanding a black-box model is a major problem in domains that relies on model predictions in critical tasks. If solved, can help to evaluate the trustworthiness of a model. This thesis proposes a user-centric approach to black-box interpretability. It addresses the problem in semantic segmentation setting with an example of humanitarian remote sensing application for building detection. The question that drives this work was, Can existing methods for explaining black-box classifiers be used for a deep learning semantic segmentation model? We approached this problem with exploratory qualitative research involving a case study and human evaluation. The study showed that it is possible to explain a segmentation model with adapted methods for classifiers but not without a cost. The specificity of the model is likely to be lost in the process. The sole process could include introducing artificial classes or fragmenting image into super-pixels. Other approaches are necessary to mitigate identified drawback. The main contribution of this work is an interactive visualisation approach for exploring learned latent space via a deep segmenter, named U-Net, evaluated with a user study involving 45 respondents. We developed an artefact (accessible online) to evaluate the approach with the survey. It presents an example of this approach with a real-world satellite image dataset. In the evaluation study, the majority of users had a computer science background (80%), including a large percentage of users with machine learning specialisation (44.4% of all respondents). The model distinguishes rurality vs urbanization (58% of users). External quantitative comparison of building densities of each city concerning the location in the latent space confirmed the later. The representation of the model was found faithful to the underlying model (62% of users). Preliminary results show the utility of the pursued approach in the application domain. Limited possibility to present complex model visually requires further investigation. / Att förstå en svartboxmodell är ett stort problem inom domäner som förlitar sig på modellprognoser i kritiska uppgifter. Om det löses, kan det hjälpa till att utvärdera en modells pålitlighet. Den här avhandlingen föreslår en användarcentrisk strategi för svartboxtolkbarhet. Den tar upp problemet i semantisk segmentering med ett exempel på humanitär fjärranalysapplikation för byggnadsdetektering. Frågan som driver detta arbete var: Kan befintliga metoder för att förklara svartruta klassificerare användas för en djup semantisk segmenteringsmodell? Vi närmade oss detta problem med utforskande kvalitativ forskning som involverade en fallstudie och mänsklig utvärdering. Studien visade att det är möjligt att förklara en segmenteringsmodell med anpassade metoder för klassificerare men inte utan kostnad. Modellens specificitet kommer sannolikt att gå förlorad i processen. Den enda processen kan inkludera införande av konstgjorda klasser eller fragmentering av bild i superpixlar. Andra tillvägagångssätt är nödvändiga för att mildra identifierad nackdel. Huvudbidraget i detta arbete är en interaktiv visualiseringsmetod för att utforska lärt latent utrymme via en djup segmenter, benämnd U-Net, utvärderad med en användarstudie med 45 svarande. Vi utvecklade en artefakt (tillgänglig online) för att utvärdera tillvägagångssättet med undersökningen. Den presenterar ett exempel på denna metod med en verklig satellitbilddatasats. I utvärderingsstudien hade majoriteten av användarna en datavetenskaplig bakgrund (80%), inklusive en stor andel användare med specialisering av maskininlärning (44,4 % av alla svarande). Modellen skiljer ruralitet och urbanisering (58 % av användarna). Den externa kvantitativa jämförelsen av byggnadstätheten i varje stad angående platsen i det latenta utrymmet bekräftade det senare. Representationen av modellen visade sig vara trogen mot den underliggande modellen (62% av användarna). Preliminära resultat visar användbarheten av den eftersträvade metoden inom applikationsdomänen. Begränsad möjlighet att presentera komplexa modeller visuellt kräver ytterligare utredning.

Online supervised learning from fast-evolving imbalanced data streams has applications in many areas. That is, the development of techniques that are able to handle highly skewed class distributions (or 'class imbalance') is an important area of research in domains such as manufacturing, the environment, and health. Solutions should be able to analyze large repositories in near real-time and provide accurate models to describe rare classes that may appear infrequently or in bursts while continuously accommodating new instances. Although numerous online learning methods have been proposed to handle binary class imbalance, solutions suitable for multi-class streams with varying degrees of imbalance in evolving streams have received limited attention. To address this knowledge gap, the first contribution of this thesis introduces the Online Learning from Imbalanced Multi-Class Streams through Dynamic Sampling (DynaQ) algorithm for learning in such multi-class imbalanced settings. Our approach utilizes a queue-based learning method that dynamically creates an instance queue for each class. The number of instances is balanced by maintaining a queue threshold and removing older samples during training. In addition, new and rare classes are dynamically added to the training process as they appear. Our experimental results confirm a noticeable improvement in minority-class detection and classification performance. A comparative evaluation shows that the DynaQ algorithm outperforms the state-of-the-art approaches. Our second contribution in this thesis focuses on fairness-aware learning from imbalanced streams. Our work is motivated by the observation that the decisions made by online learning algorithms may negatively impact individuals or communities. Indeed, the development of approaches to handle these concerns is an active area of research in the machine learning community. However, most existing methods process the data in offline settings and are not directly suitable for online learning from evolving data streams. Further, these techniques fail to take the effects of class imbalance, on fairness-aware supervised learning into account. In addition, recent fairness-aware online learning supervised learning approaches focus on one sensitive attribute only, which may lead to subgroup discrimination. In a fair classification, the equality of fairness metrics across multiple overlapping groups must be considered simultaneously. In our second contribution, we thus address the combined problem of fairness-aware online learning from imbalanced evolving streams, while considering multiple sensitive attributes. To this end, we introduce the Multi-Sensitive Queue-based Online Fair Learning (MQ-OFL) algorithm, an online fairness-aware approach, which maintains valid and fair models over evolving streams. MQ-OFL changes the training distribution in an online fashion based on both stream imbalance and discriminatory behavior of the model evaluated over the historical stream. We compare our MQ-OFL method with state-of-art studies on real-world datasets and present comparative insights on the performance. Our final contribution focuses on explainability and interpretability in fairness-aware online learning. This research is guided by the concerns raised due to the black-box nature of models, concealing internal logic from users. This lack of transparency poses practical and ethical challenges, particularly when these algorithms make decisions in finance, healthcare, and marketing domains. These systems may introduce biases and prejudices during the learning phase by utilizing complex machine learning algorithms and sensitive data. Consequently, decision models trained on such data may make unfair decisions and it is important to realize such issues before deploying the models. To address this issue, we introduce techniques for interpreting the outcomes of fairness-aware online learning. Through a case study predicting income based on features such as ethnicity, biological sex, age, and education level, we demonstrate how our fairness-aware learning process (MQ-OFL) maintains a balance between accuracy and discrimination trade-off using global and local surrogate models.

Machine learning has revolutionized recommendation systems by employing ranking models for personalized item suggestions. However, the complexity of learning-to-rank (LTR) models poses challenges in understanding the underlying reasons contributing to the ranking outcomes. This lack of transparency raises concerns about potential errors, biases, and ethical implications. To address these issues, interpretable LTR models have emerged as a solution. Currently, the state-of-the-art for interpretable LTR models is led by generalized additive models (GAMs). However, ranking GAMs face limitations in terms of computational intensity and handling high-dimensional data. To overcome these drawbacks, post-hoc methods, including local interpretable modelagnostic explanations (LIME), have been proposed as potential alternatives. Nevertheless, a quantitative evaluation comparing post-hoc methods efficacy to state-of-the-art ranking GAMs remains largely unexplored. This study aims to investigate the capabilities and limitations of LIME in an attempt to approximate a complex ranking model using a surrogate model. The proposed methodology for this study is an experimental approach. The neural ranking GAM, trained on two benchmark information retrieval datasets, serves as the ground truth for evaluating LIME’s performance. The study adapts LIME in the context of ranking by translating the problem into a classification task and asses three different sampling strategies against the prevalence of imbalanced data and their influence on the correctness of LIME’s explanations. The findings of this study contribute to understanding the limitations of LIME in the context of ranking. It analyzes the low similarity between the explanations of LIME and those generated by the ranking model, highlighting the need to develop more robust sampling strategies specific to ranking. Additionally, the study emphasizes the importance of developing appropriate evaluation metrics for assessing the quality of explanations in ranking tasks.

There is an increasing interest in using deep neural networks in various downstream natural language processing tasks. Such models are commonly used as black boxes, meaning that their decision-making is difficult to interpret. In order to build trust in models, it is crucial to analyse their inner workings which lead to predictions. The need to interpret natural language processing models has induced research on linguistically-informed interpretability. This field revolves around choosing specific linguistic phenomena and inspecting models' capability to capture them without being explicitly trained for it.  This thesis project contributes to the field by assessing the ability of BERT-style models to learn subject number in Lithuanian and English. The experiments revolve around designing diagnostic classifiers which are used to determine if the models are capable of learning this particular linguistic phenomenon. The results show that BERT-style models are capable of implicitly learning the number of a subject both in Lithuanian and English. However, this seems to be harder in Lithuanian, as diagnostic classifiers show a lower accuracy. The study observes that the accuracy of logistic regression diagnostic classifiers fluctuates to a large extent. Fully connected neural network classifiers outperform logistic regression classifiers.

BACKGROUND. Multimodal Machine Learning is a powerful paradigm that capitalizes on the complementary predictive capabilities of different data modalities, such as text, image, time series. This approach allows for an extremely diverse feature space, which proves useful for combining different real-world tasks into a single model. Current architectures in the field of multimodal learning often integrate feature representations in parallel, a practice that not only limits their interpretability but also creates a reliance on the availability of specific modalities. Interpretability and robustness to missing inputs are particularly important in clinical decision support systems. To address these issues, the iGH Research Group at EPFL proposed a modular sequential input fusion called Modular Decision Support Network (MoDN). MoDN was tested on unimodal tabular inputs for multitask outputs and was shown to be superior to its monolithic parallel counterparts, while handling any number and combination of inputs and providing continuous real-time predictive feedback. AIM. We aim to extend MoDN to MultiModN with multimodal inputs and compare the benefits and limitations of sequential fusion with a state-of-the-art parallel fusion (P-Fusion) baseline.METHODS & FINDINGS. We align our experimental setup with a previously published P-Fusion baseline, focusing on two binary diagnostic predictive tasks (presence of pleural effusion and edema) in a popular multimodal clinical benchmark dataset (MIMIC).We perform four experiments: 1) comparing MultiModN to P-Fusion, 2) extending the architecture to multiple tasks, 3) exploring MultiModN's inherent interpretability in several metrics, and 4) testing its ability to be resistant to biased missingness by simulating missing not at random (MNAR) data during training and flipping the bias at inference. We show that MultiModN's sequential architecture does not compromise performance compared with the P-Fusion baseline, despite the added advantages of being multitask, composable and inherently interpretable. The final experiment shows that MultiModN resists catastrophic failure from MNAR data, which is particularly prevalent in clinical settings. / Multimodal maskininlärning är ett kraftfullt paradigm som utnyttjar de kompletterande prediktiva egenskaperna hos olika datamodaliteter, såsom text, bild, tidsserier. Detta tillvägagångssätt möjliggör ett extremt varierat funktionsutrymme, vilket visar sig vara användbart för att kombinera olika verkliga uppgifter i en enda modell. Nuvarande arkitekturer för multimodal inlärning integrerar ofta funktionsrepresentationer parallellt, en praxis som inte bara begränsar deras tolkningsbarhet utan också skapar ett beroende av tillgängligheten av specifika modaliteter. Tolkningsbarhet och robusthet mot saknade indata är särskilt viktigt i kliniska beslutsstödsystem. För att lösa dessa problem har forskargruppen iGH vid EPFL föreslagit en modulär sekventiell fusion av indata som kallas Modular Decision Support Network (MoDN). MoDN testades på unimodala tabulära indata för multitask-utdata och visade sig vara överlägsen sina monolitiska parallella motsvarigheter, samtidigt som den hanterar alla antal och kombinationer av indata och ger kontinuerlig prediktiv feedback i realtid. Vårt mål är att utöka MoDN till MultiModN med multimodala indata och jämföra fördelarna och begränsningarna med sekventiell fusion med en toppmodern baslinje för parallell fusion (P-Fusion). Vi anpassar vår experimentuppsättning till en tidigare publicerad P-Fusion-baslinje, med fokus på två binära diagnostiska prediktiva uppgifter (närvaro av pleural effusion och ödem) i en populär multimodal klinisk benchmark datauppsättning (MIMIC), som omfattar bilder, text, tabelldata och tidsserier. Vi utför fyra experiment och visar att MultiModN:s sekventiella arkitektur inte försämrar prestandan jämfört med P-Fusions baslinje, trots de extra fördelarna med att vara multitasking, komponerbar och tolkningsbar i sin egen rätt. Det sista experimentet visar att MultiModN motstår katastrofala fel från MNAR-data, vilket är särskilt vanligt i kliniska miljöer.

Insolvenzprognosen und Ratings sind wichtige Aufgaben der Finanzbranche und dienen der Kreditwürdigkeitsprüfung von Unternehmen. Eine Möglichkeit dieses Aufgabenfeld anzugehen, ist maschinelles Lernen. Dabei werden Vorhersagemodelle aufgrund von Beispieldaten aufgestellt. Methoden aus diesem Bereich sind aufgrund Ihrer Automatisierbarkeit vorteilhaft. Dies macht menschliche Expertise in den meisten Fällen überflüssig und bietet dadurch einen höheren Grad an Objektivität. Allerdings sind auch diese Ansätze nicht perfekt und können deshalb menschliche Expertise nicht gänzlich ersetzen. Sie bieten sich aber als Entscheidungshilfen an und können als solche von Experten genutzt werden, weshalb interpretierbare Modelle wünschenswert sind. Leider bieten nur wenige Lernalgorithmen interpretierbare Modelle. Darüber hinaus sind einige Aufgaben wie z.B. Rating häufig Mehrklassenprobleme. Mehrklassenklassifikationen werden häufig durch Meta-Algorithmen erreicht, welche mehrere binäre Algorithmen trainieren. Die meisten der üblicherweise verwendeten Meta-Algorithmen eliminieren jedoch eine gegebenenfalls vorhandene Interpretierbarkeit. In dieser Dissertation untersuchen wir die Vorhersagegenauigkeit von interpretierbaren Modellen im Vergleich zu nicht interpretierbaren Modellen für Insolvenzprognosen und Ratings. Wir verwenden disjunktive Normalformen und Entscheidungsbäume mit Schwellwerten von Finanzkennzahlen als interpretierbare Modelle. Als nicht interpretierbare Modelle werden Random Forests, künstliche Neuronale Netze und Support Vector Machines verwendet. Darüber hinaus haben wir einen eigenen Lernalgorithmus Thresholder entwickelt, welcher disjunktive Normalformen und interpretierbare Mehrklassenmodelle generiert. Für die Aufgabe der Insolvenzprognose zeigen wir, dass interpretierbare Modelle den nicht interpretierbaren Modellen nicht unterlegen sind. Dazu wird in einer ersten Fallstudie eine in der Praxis verwendete Datenbank mit Jahresabschlüssen von 5152 Unternehmen verwendet, um die Vorhersagegenauigkeit aller oben genannter Modelle zu messen. In einer zweiten Fallstudie zur Vorhersage von Ratings demonstrieren wir, dass interpretierbare Modelle den nicht interpretierbaren Modellen sogar überlegen sind. Die Vorhersagegenauigkeit aller Modelle wird anhand von drei in der Praxis verwendeten Datensätzen bestimmt, welche jeweils drei Ratingklassen aufweisen. In den Fallstudien vergleichen wir verschiedene interpretierbare Ansätze bezüglich deren Modellgrößen und der Form der Interpretierbarkeit. Wir präsentieren exemplarische Modelle, welche auf den entsprechenden Datensätzen basieren und bieten dafür Interpretationsansätze an. Unsere Ergebnisse zeigen, dass interpretierbare, schwellwertbasierte Modelle den Klassifikationsproblemen in der Finanzbranche angemessen sind. In diesem Bereich sind sie komplexeren Modellen, wie z.B. den Support Vector Machines, nicht unterlegen. Unser Algorithmus Thresholder erzeugt die kleinsten Modelle während seine Vorhersagegenauigkeit vergleichbar mit den anderen interpretierbaren Modellen bleibt. In unserer Fallstudie zu Rating liefern die interpretierbaren Modelle deutlich bessere Ergebnisse als bei der zur Insolvenzprognose (s. o.). Eine mögliche Erklärung dieser Ergebnisse bietet die Tatsache, dass Ratings im Gegensatz zu Insolvenzen menschengemacht sind. Das bedeutet, dass Ratings auf Entscheidungen von Menschen beruhen, welche in interpretierbaren Regeln, z.B. logischen Verknüpfungen von Schwellwerten, denken. Daher gehen wir davon aus, dass interpretierbare Modelle zu den Problemstellungen passen und diese interpretierbaren Regeln erkennen und abbilden.

L'objectif de cette thèse est d'améliorer la recherche d'objets visuels à l'aide de l'interactivité avec l'utilisateur. Notre solution est de construire un système intéractif permettant aux utilisateurs de définir leurs propres concepts visuels à partir de certains mots-clés visuels. Ces mots-clés visuels, qui en théorie représentent les mots visuels les plus informatifs liés à une catégorie d'objets, sont appris auparavant à l'aide d'un algorithme d'apprentissage supervisé et d'une manière discriminative. Le challenge est de construire des mots-clés visuels concis et interprétables. Notre contribution repose sur deux points. D'abord, contrairement aux approches existantes qui utilisent les sacs de mots, nous proposons d'employer les descripteurs locaux sans aucune quantification préalable. Deuxièmement, nous proposons d'ajouter une contrainte de régularisation à la fonction de perte de notre classifieur pour favoriser la parcimonie des modèles produits. La parcimonie est en effet préférable pour sa concision (nombre de mots visuels réduits) ainsi pour sa diminution du temps de prédiction. Afin d'atteindre ces objectifs, nous avons développé une méthode d'apprentissage à instances multiples utilisant une version modifiée de l'algorithme BLasso. Cet algorithme est une forme de boosting qui se comporte similairement au LASSO (Least Absolute Shrinkage and Selection Operator). Il régularise efficacement la fonction de perte avec une contrainte additive de type L1 et ceci en alternant entre des itérations en avant et en arrière. La méthode proposée est générique dans le sens où elle pourrait être utilisée avec divers descripteurs locaux voire un ensemble structuré de descripteurs locaux qui décrit une région locale de l'image. / This thesis is an attempt to improve visual object retrieval by allowing users to interact with the system. Our solution lies in constructing an interactive system that allows users to define their own visual concept from a concise set of visual patches given as input. These patches, which represent the most informative clues of a given visual category, are trained beforehand with a supervised learning algorithm in a discriminative manner. Then, and in order to specialize their models, users have the possibility to send their feedback on the model itself by choosing and weighting the patches they are confident of. The real challenge consists in how to generate concise and visually interpretable models. Our contribution relies on two points. First, in contrast to the state-of-the-art approaches that use bag-of-words, we propose embedding local visual features without any quantization, which means that each component of the high-dimensional feature vectors used to describe an image is associated to a unique and precisely localized image patch. Second, we suggest using regularization constraints in the loss function of our classifier to favor sparsity in the models produced. Sparsity is indeed preferable for concision (a reduced number of patches in the model) as well as for decreasing prediction time. To meet these objectives, we developed a multiple-instance learning scheme using a modified version of the BLasso algorithm. BLasso is a boosting-like procedure that behaves in the same way as Lasso (Least Absolute Shrinkage and Selection Operator). It efficiently regularizes the loss function with an additive L1-constraint by alternating between forward and backward steps at each iteration. The method we propose here is generic in the sense that it can be used with any local features or feature sets representing the content of an image region. / تعالج هذه الأطروحة مسألة البحث عن الأشياء في الصور الثابتة و هي محاولة لتحسين نتائج البحث المنتظرة عن طريق تفاعل المستخدم مع النظام . يتمثل الحل المقترح في تصميم نظام تفاعلي يتيح للمستخدم صياغة مفهومه المرئي عن طريق مجموعة مقتضبة من أجزاء صغيرة للصور هي عبارة عن كلمات مفاتيح قد تم تعلمها سابقا عن طريق تعلم آلي استنتاجي . يمكن للمستخدم حينئذ تخصيص أنموذجه أولا بالاختيار ثم بترجيح الأجزاء التي يراها مناسبة . يتمثل التحدي القائم في كيفية توليد نماذج مرئية مفهومة و مقتضبة . نكون قد ساهمنا في هذا المجال بنقطتين أساسيتين تتمثل الأولى في إدماج الواصفات المحلية للصور دون أي تكميم ، و بذلك يكون كل مكون من ناقلات الميزات ذات الأبعاد العالية مرتبط حصريا بمكان وحيد و محدد في الصورة . ثانيا ، نقترح إضافة قيود تسوية لدالة الخسارة من أجل التحصل على حلول متفرقة و مقتضبة . يساهم ذلك في تقلص عدد هذه الأجزاء المرئية و بالتالي في ربح إضافي لوقت التكهن . في إطار تحقيق الأهداف المرسومة ، قمنا بإعداد مشروع تعلم قائم على تعدد الأمثلة يرتكز أساسا على نسخة محورة لخوارزمية بلاسو . تجدر الإشارة في الأخير أنه يمكن توظيف هذا العمل باستخدام نوع أو عدة أنواع من الواصفات المحلية للصور.

A Escala Facial de Dor - Revisada (EFD-R) é uma das escalas mais recomendadas na mensuração da intensidade da dor aguda em crianças. A versão original desta escala foi testada em crianças canadenses. O objetivo deste trabalho foi avaliar a validade, interpretação e preferência da versão brasileira da Escala Facial de Dor - Revisada (EFD-R-B), em duas amostras de crianças brasileiras: uma envolvendo dor aguda procedural e outra dor aguda pós-cirúrgica. Na primeira amostra participaram 77 crianças com idades entre 6 e 12 anos, do sexo feminino e masculino, que foram submetidas à coleta de sangue (dor procedural). As crianças estimaram a intensidade da sua dor, antes e após a punção venosa, na EFD-R-B. Na estimação após a punção venosa, a Escala Colorida Analógica (ECA) foi administrada junto com a EFD-R-B e, além disso, as crianças indicaram as faces que expressavam uma dor leve, moderada e severa, a escala que preferiam e o porquê. Na segunda amostra, participaram 53 crianças com idades entre 6 e 12 anos, do sexo feminino e masculino, que tinham sido submetidas a pequenas cirurgias (dor pós-cirúrgica). Nesta amostra, as crianças estimaram, na EFD-R-B e na ECA, a intensidade da dor que estavam sentindo no momento da entrevista. Também indicaram as faces que expressavam uma dor leve, moderada e severa, o limiar de tratamento da dor, a escala que preferiam e o porquê. Na comparação entre as pontuações obtidas na EFD-R-B e na ECA (validade convergente), nas duas amostras, os valores dos coeficientes Kendall\'s tau foram altos e significativos: =0,75 para o grupo de dor procedural e =0,79 para o grupo de dor pós-cirúrgica (p=0,00 nas duas amostras). No grupo de dor procedural, a EFD-R-B refletiu as mudanças na intensidade da dor vivenciada pelas crianças antes e após a punção venosa (validade concorrente): Teste de Wilcoxon z=-6,65; p=0,00. Considerando uma escala de 0 a 10 para a EFD-R-B, a mediana e a amplitude interquartil (AIQ) para as faces indicadas como expressivas de intensidade leve, moderada e severa, foram 2 (2-2), 4 (4-6) e 10 (10-10) respectivamente, no grupo de dor procedural, e 2 (2-2), 6 (4-8) e 10 (10-10) respectivamente, no grupo de dor pós-cirúrgica. Na estimação do limiar de tratamento da dor (grupo de dor pós-cirúrgica), a mediana (AIQ) foi 6 (4-10). No grupo de dor procedural, a EFD-R-B foi a escala preferida por 57,1% das crianças e a ECA por 41,6%; no grupo de dor pós-cirúrgica, a EFD-R-B foi escolhida por 66% das crianças e a ECA por 34%. Estas proporções somente foram significativas no grupo de dor pós-cirúrgica (X²=5,453 p=0,02). Nossos resultados mostram que a EFD-R-B possui propriedades similares à escala original e boa aceitação entre as crianças entrevistadas. A determinação dos valores das diferentes intensidades de dor e do limiar de tratamento da dor, para cada participante, representa uma evidência importante sobre a interpretação da EFD-R. / The Faces Pain Scale Revised (FPS-R) is one of the most recommended tools in measuring the intensity of acute pain in children. The aim of this study was to assess validity, interpretability and preference of the Brazilian version of the FPS-R (FPS-R-B), in two different clinical samples. The first sample contained seventy-seven children, 6 to 12 years old and both sexes, undergoing venipuncture for blood sample (procedural pain). These children estimated their perceived pain intensity in FPS-R-B before and after venipuncture. Furthermore, after venipuncture, children were asked: a) to evaluate the intensity of their needle pain using the Coloured Analogue Scale (CAS), b) to indicate on the Faces scale the intensities representing the mild, moderate and severe pain, and c) to choose the scale they preferred and indicate the reasons for the preference. The second sample included fifty-three children, 6 to 12 years old and both sexes, undergoing minor surgery (postoperative pain). Following surgery, children were asked: a) to provide a rating of their current pain intensity using the FPS-R-B and the CAS, b) to indicate on the Faces scale the intensities representing the mild, moderate and severe pain, c) to estimate, on the FPS-R-B, the intensity of pain that their felt to warrant pharmacologic intervention (pain treatment threshold), and d) to choose the scale they preferred and indicate the reasons for the preference. The degree of concordance between FPS-R-B and CAS ratings (convergent validity), for both samples, was high and statistically significant Kendall\'s tau value was 0.75 for the first sample, and 0.79 for the second sample, (p<0.05) . FPS-R-B reflected the changes in pain intensity before and after venipuncture (concurrent validity): Wilcoxon Test z=- 6.24; p< 0.05. On the 0-10 scale for the FPS-R-B, the median and interquartile range (IQR) of the intensities that represented mild, moderate and severe pain were 2 (2-2), 4 (4-6) e 10 (10-10) respectively, for the first sample, and 2 (2-2), 6 (4-8) e 10 (10-10) respectively, for the second sample. The median and IQR for pain treatment threshold were 6 (4-10). Fifty-seven percent of children in the first sample and 64.8% in the second sample preferred the FPS-R-B. These proportions were statistically significant for the second sample (X²=5,453 p<0,05). Our data show that the FPS-R-B has similar statistical properties to the original. New evidences were presented regarding interpretability of the FPS-R by determining each children\'s treatment threshold and estimate of mild, moderate and severe pain. In this study, the FPS-R-B was preferred by the majority of children.

Esta dissertação apresenta um sistema de indução de classificadores fuzzy. Ao invés de utilizar a abordagem tradicional de sistemas fuzzy baseados em regras, foi utilizado o modelo de Árvore de Padrões Fuzzy(APF), que é um modelo hierárquico, com uma estrutura baseada em árvores que possuem como nós internos operadores lógicos fuzzy e as folhas são compostas pela associação de termos fuzzy com os atributos de entrada. O classificador foi obtido sintetizando uma árvore para cada classe, esta árvore será uma descrição lógica da classe o que permite analisar e interpretar como é feita a classificação. O método de aprendizado originalmente concebido para a APF foi substituído pela Programação Genética Cartesiana com o intuito de explorar melhor o espaço de busca. O classificador APF foi comparado com as Máquinas de Vetores de Suporte, K-Vizinhos mais próximos, florestas aleatórias e outros métodos Fuzzy-Genéticos em diversas bases de dados do UCI Machine Learning Repository e observou-se que o classificador APF apresenta resultados competitivos. Ele também foi comparado com o método de aprendizado original e obteve resultados comparáveis com árvores mais compactas e com um menor número de avaliações. / This work presents a system for induction of fuzzy classifiers. Instead of the traditional fuzzy based rules, it was used a model called Fuzzy Pattern Trees (FPT), which is a hierarchical tree-based model, having as internal nodes, fuzzy logical operators and the leaves are composed of a combination of fuzzy terms with the input attributes. The classifier was obtained by creating a tree for each class, this tree will be a logic class description which allows the interpretation of the results. The learning method originally designed for FPT was replaced by Cartesian Genetic Programming in order to provide a better exploration of the search space. The FPT classifier was compared against Support Vector Machines, K Nearest Neighbour, Random Forests and others Fuzzy-Genetics methods on several datasets from the UCI Machine Learning Repository and it presented competitive results. It was also compared with Fuzzy Pattern trees generated by the former learning method and presented comparable results with smaller trees and a lower number of functions evaluations.

Made available in DSpace on 2016-06-02T19:05:32Z (GMT). No. of bitstreams: 1 2113.pdf: 882226 bytes, checksum: 022c380c1d469988d9e4617a030f17c3 (MD5) Previous issue date: 2008-08-06 / The combination of fuzzy systems with clustering algorithms has great acceptance in the scientific community mainly due to its adherence to the advantage balance principle of computational intelligence, in which different methodologies collaborate with each other potentializing the usefulness and applicability of the resulting systems. Fuzzy Modeling using clustering algorithms presents the transparency and comprehensibility typical of the linguistic fuzzy systems at the same time that benefits from the possibilities of dimensionality reduction by means of clustering. In this work is presented the Fuzzy-CCM method (Fuzzy Conditional Clustering based Modeling) which consists of a new approach for Fuzzy Modeling based on the Fuzzy Conditional Clustering algorithm aiming at providing new means to address the topic of interpretability of fuzzy rules bases. With the Fuzzy-CCM method the balance between interpretability and accuracy of fuzzy rules is dealt with through the definition of contexts defined by a small number of input variables and the generation of clusters induced by these contexts. The rules are generated in a different format, with linguistic variables and clusters in the antecedent. Some experiments have been carried out using different knowledge domains in order to validate the proposed approach by comparing the results with the ones obtained by the Wang&Mendel and conventional Fuzzy C-Means methods. The theoretical foundations, the advantages of the method, the experiments and results are presented and discussed. / A combinação de sistemas fuzzy com algoritmos de agrupamento tem grande aceitação na comunidade científica devido; principalmente, a sua aderência ao princípio de balanceamento de vantagens da inteligência computacional, no qual metodologias diferentes colaboram entre si, potencializando a utilidade e aplicabilidade dos sistemas resultantes. A modelagem fuzzy usando algoritmos de agrupamento apresenta a transparência e facilidade de compreensão típica dos sistemas fuzzy lingüísticos ao mesmo tempo em que se beneficia das possibilidades de redução da dimensionalidade por intermédio do agrupamento. Neste trabalho é apresentado o método Fuzzy-CCM (Fuzzy Conditional Clustering based Modeling), que consiste de uma nova abordagem de Modelagem Fuzzy baseada no algoritmo de Agrupamento Fuzzy Condicional, cujo objetivo é prover novos meios de tratar a questão da interpretabilidade de bases de regras fuzzy. Com o método Fuzzy-CCM, o balanço entre interpretabilidade e acuidade de regras fuzzy é tratado por meio da definição de contextos formados com um pequeno número de variáveis de entrada e a geração de grupos condicionados por estes contextos. As regras são geradas em um formato diferente, que contêm variáveis lingüísticas e grupos no seu antecedente. Alguns experimentos foram executados usando diferentes domínios de conhecimento a fim de validar a abordagem proposta, comparando os resultados obtidos usando a nova abordagem com os resultados obtidos usando os métodos Wang&Mendel e Fuzzy C-Means. A fundamentação teórica, as vantagens do método, os experimentos e os resultados obtidos são apresentados e discutidos.