Ao se realizar estudo em qualquer área do conhecimento, quanto mais dados se dispuser, maior a dificuldade de se extrair conhecimento útil deste banco de dados. A finalidade deste trabalho é apresentar algumas ferramentas ditas inteligentes, de extração de conhecimento destes grandes repositórios de dados. Apesar de ter várias conotações, neste trabalho, irá se entender extração de conhecimento dos repositórios de dados a ocorrência combinada de alguns dados com freqüência e confiabilidade que se consideram interessantes, ou seja, na medida e que determinado dado ou conjunto de dados aparece no repositório de dados, em freqüência considerada razoável, outro dado ou conjunto de dados irá aparecer. Executada sobre repositórios de dados referentes a informações georreferenciadas dos alunos da UERJ (Universidade do Estado do Rio de Janeiro), irá se analisar os resultados de duas ferramentas de extração de dados, bem como apresentar possibilidades de otimização computacional destas ferramentas. / Comparative Study of Techniques for Extracting knowledge from large data repositories. When conducting the study in any field of knowledge, the more data is available, the greater the difficulty in extracting useful knowledge from this database. The purpose of this paper is to present some tools called intelligent, knowledge extraction of these large data repositories. Although many connotations, this work will understand knowledge extraction from data repositories on the combined occurrence of some data with frequency and reliability that are considered interesting, ie, the extent and specific data or data set appears in the data, at a rate deemed reasonable, other data or data set will appear. Runs on repositories of data on georeferenced data of students UERJ (Universidade do Estado do Rio de Janeiro), will analyze the results of two tools to extract data and present opportunities for optimization of these computational tools.

Ao se realizar estudo em qualquer área do conhecimento, quanto mais dados se dispuser, maior a dificuldade de se extrair conhecimento útil deste banco de dados. A finalidade deste trabalho é apresentar algumas ferramentas ditas inteligentes, de extração de conhecimento destes grandes repositórios de dados. Apesar de ter várias conotações, neste trabalho, irá se entender extração de conhecimento dos repositórios de dados a ocorrência combinada de alguns dados com freqüência e confiabilidade que se consideram interessantes, ou seja, na medida e que determinado dado ou conjunto de dados aparece no repositório de dados, em freqüência considerada razoável, outro dado ou conjunto de dados irá aparecer. Executada sobre repositórios de dados referentes a informações georreferenciadas dos alunos da UERJ (Universidade do Estado do Rio de Janeiro), irá se analisar os resultados de duas ferramentas de extração de dados, bem como apresentar possibilidades de otimização computacional destas ferramentas. / Comparative Study of Techniques for Extracting knowledge from large data repositories. When conducting the study in any field of knowledge, the more data is available, the greater the difficulty in extracting useful knowledge from this database. The purpose of this paper is to present some tools called intelligent, knowledge extraction of these large data repositories. Although many connotations, this work will understand knowledge extraction from data repositories on the combined occurrence of some data with frequency and reliability that are considered interesting, ie, the extent and specific data or data set appears in the data, at a rate deemed reasonable, other data or data set will appear. Runs on repositories of data on georeferenced data of students UERJ (Universidade do Estado do Rio de Janeiro), will analyze the results of two tools to extract data and present opportunities for optimization of these computational tools.

This Master’s thesis deals with adaptation of fuzzy system parameters with main aim on artificial neural network. Current knowledge of methods connecting fuzzy systems and artificial neural networks is discussed in the search part of this work. The search in Student’s works is discussed either. Chapter focused on methods application deals with classifying ability verification of the chosen fuzzy-neural network with Kohonen learning algorithm. Later the model of fuzzy system with parameters adaptation based on fuzzyneural network with Kohonen learning algorithm is shown.

This diploma thesis deals with development of a fuzzy decision-making system for evaluating of REONTECH CZ s.r.o. company suppliers. The first part is dedicated to theory of fuzzy logic. The next part contains firstly an analysis of the current situation and then a description of the development process of the fuzzy decision-making system. The main outcome of this thesis is a tool that meets needs and requirements of the company and serves as a decision-making support tool.

This dissertation is a systematic study of artificial intelligence (AI) applications for the diagnosis of power transformer incipient fault. The AI techniques include artificial neural networks (ANN, or briefly neural networks - NN), expert systems, fuzzy systems and multivariate regression. The fault diagnosis is based on dissolved gas-in-oil analysis (DGA). A literature review showed that the conventional fault diagnosis methods, i.e. the ratio methods (Rogers, Dornenburg and IEC) and the key gas method, have limitations such as the "no decision" problem. Various AI techniques may help solve the problems and present a better solution. Based on the IEC 599 standard and industrial experiences, a knowledge-based inference engine for fault detection was developed. Using historical transformer failure data from an industrial partner, a multi-layer perceptron (MLP) modular neural network was identified as the best choice among several neural network architectures. Subsequently, the concept of a hybrid diagnosis was proposed and implemented, resulting in a combined neural network and expert system tool (the ANNEPS system) for power transformer incipient diagnosis. The abnormal condition screening process, as well as the principle and algorithms of combining the outputs of knowledge based and neural network based diagnosis, were proposed and implemented in the ANNEPS. Methods of fuzzy logic based transformer oil/paper insulation condition assessment, and estimation of oil sampling interval and maintenance recommendations, were also proposed and implemented. Several methods of power transformer incipient fault location were investigated, and a 7Ã 21Ã 5 MLP network was identified as the best choice. Several methods for on-load tap changer (OLTC) coking diagnosis were also investigated, and a MLP based modular network was identified as the best choice. Logistic regression analysis was identified as a good auditor in neural network input pattern selection processes. The above results can help developing better power transformer maintenance strategies, and serve as the basis of on-line DGA transformer monitors. / Ph. D.

Submitted by Maria Aparecida (cidazen@gmail.com) on 2017-04-17T12:44:32Z No. of bitstreams: 1 Hugo Alves Velozo.pdf: 5196080 bytes, checksum: 14e9edcc07c0256cf726b1d0f7eb9a02 (MD5) / Made available in DSpace on 2017-04-17T12:44:32Z (GMT). No. of bitstreams: 1 Hugo Alves Velozo.pdf: 5196080 bytes, checksum: 14e9edcc07c0256cf726b1d0f7eb9a02 (MD5) Previous issue date: 2017-02-20 / CAPES / Most methods of identifcation of closed-loop dynamic systems are developed for linear and deterministic systems. However, most closed loop systems are nonlinear dynamic systems. In addition, such systems are subject to stochastic perturbations. Considering this problem, this work presents a methodology for the identifcation of closed loop stochastic nonlinear systems. For this purpose, the proposed methodology uses a local approach to identify nonlinear dynamic systems, that is, a set of Box-Jenkins local models are used to identify the dynamics of the nonlinear system. In this work, the nonlinear system is modeled through a Takagi-Sugeno fuzzy inference system, where the parameters of the antecedent of the fuzzy rules are estimated with the fuzzy clustering algorithm GustafsonKessel and the consequent Box-Jenkins model parameters are estimated with the fuzzy fuzzy RIV (Refned Instrumental Variable) and fuzzy IVARMA (Instrumental Variable ARMA) algorithms. The proposed method is applied in the identifcation of a closed-loop nonlinear thermal plant. / A maioria dos métodos de identifcação de sistemas dinâmicos em malha fechada são desenvolvidos para sistemas lineares e determinísticos. Entretanto, a maioria dos sistemas operando em malha fechada são sistemas dinâmicos não lineares. Além disso, esses sistemas estão sujeitos a perturbações de natureza estocástica. Considerando essa problemática, este trabalho apresenta uma metodologia para identifcação de sistemas não lineares estocásticos em malha fechada. Para isso, a metodologia proposta utiliza uma abordagem local de identifcação de sistemas dinâmicos não lineares, ou seja, um conjunto de modelos locais Box-Jenkins são utilizados para identifcar a dinâmica do sistema não linear. Neste trabalho, o sistema não linear é modelado por meio de um sistema de inferência fuzzy Takagi-Sugeno, onde os parâmetros do antecedente das regras fuzzy são estimados com o algoritmo de agrupamento fuzzy Gustafson-Kessel e o parâmetros do modelo Box-Jenkins do consequente são estimados com os algoritmos RIV (Refned Instrumental Variable) fuzzy e IVARMA (Instrumental Variable ARMA) fuzzy. O método proposto é aplicado na identifcação de uma planta térmica não linear em malha fechada.

We propose a novel approach for building a type-2 neuro-fuzzy system from a given set of input-output training data. For an input pattern, a corresponding crisp output of the system is obtained by combining the inferred results of all the rules into a type-2 fuzzy set which is then defuzzified by applying a type reduction algorithm. Karnik and Mendel proposed an algorithm, called KM algorithm, to compute the centroid of an interval type-2 fuzzy set efficiently. Based on this algorithm, Liu developed a centroid type-reduction strategy to do type reduction for type-2 fuzzy sets. A type-2 fuzzy set is decomposed into a collection of interval type-2 fuzzy sets by £\-cuts. Then the KM algorithm is called for each interval type-2 fuzzy set iteratively. However, the initialization of the switch point in each application of the KM algorithm is not a good one. In this thesis, we present an improvement to Liu's algorithm. We employ the result previously obtained to construct the starting values in the current application of the KM algorithm. Convergence in each iteration except the first one can then speed up and type reduction for type-2 fuzzy sets can be done faster. The efficiency of the improved algorithm is analyzed mathematically and demonstrated by experimental results. Constructing a type-2 neuro-fuzzy system involves two major phases, structure identification and parameter identification. We propose a method which incorporates self-constructing fuzzy clustering algorithm and a SVD-based least squares estimator for structure identification of type-2 neuro-fuzzy modeling. The self-constructing fuzzy clustering method is used to partition the training data set into clusters through input-similarity and output-similarity tests. The membership function associated with each cluster is defined with the mean and deviation of the data points included in the cluster. Then applying SVD-based least squares estimator, a type-2 fuzzy TSK IF-THEN rule is derived from each cluster to form a fuzzy rule base. After that a fuzzy neural network is constructed. In the parameter identification phase, the parameters associated with the rules are then refined through learning. We propose a hybrid learning algorithm which incorporates particle swarm optimization and a SVD-based least squares estimator to refine the antecedent parameters and the consequent parameters, respectively. We demonstrate the effectiveness of our proposed approach in constructing type-2 neuro-fuzzy systems by showing the results for two nonlinear functions and two real-world benchmark datasets. Besides, we use the proposed approach to construct a type-2 neuro-fuzzy system to forecast the daily Taiwan Stock Exchange Capitalization Weighted Stock Index (TAIEX). Experimental results show that our forecasting system performs better than other methods.

Neuro-fuzzy modeling is an efficient computing paradigm for system modeling problems. It mainly integrates two well-known approaches, neural networks and fuzzy systems, and therefore possesses advantages of them, i.e., learning capability, robustness, human-like reasoning, and high understandability. Up to now, many approaches have been proposed for neuro-fuzzy modeling. However, it still exists many problems need to be solved. We propose in this thesis two self-constructing rule generation methods, i.e., similarity-based rule generation (SRG) and similarity-and-merge-based rule generation (SMRG), and one hybrid learning algorithm (HLA) for structure identification and parameter identification, respectively, of neuro-fuzzy modeling. SRG and SMRG group the input-output training data into a set of fuzzy clusters incrementally based on similarity tests on the input and output spaces. Membership functions associated with each cluster are defined according to statistical means and deviations of the data points included in the cluster. Additionally, SMRG employs a merging mechanism to merge similar clusters dynamically. Then a zero-order or first-order TSK-type fuzzy IF-THEN rule is extracted from each cluster to form an initial fuzzy rule-base which can be directly employed for fuzzy reasoning or be further refined in the next phase of parameter identification. Compared with other methods, both our SRG and SMRG have advantages of generating fuzzy rules quickly, matching membership functions closely with the real distribution of the training data points, and avoiding the generation of the whole set of clusters from the scratch when new training data are considered. Besides, SMRG supports a more reasonable and quick mechanism for cluster merging to alleviate the problems of data-input-order bias and redundant clusters, which are encountered in SRG and other incremental clustering approaches. To refine the fuzzy rules obtained in the structure identification phase, a zero-order or first-order TSK-type fuzzy neural network is constructed accordingly in the parameter identification phase. Then, we develop a HLA composed by a recursive SVD-based least squares estimator and the gradient descent method to train the network. Our HLA has the advantage of alleviating the local minimal problem. Besides, it learns faster, consumes less memory, and produces lower approximation errors than other methods. To verify the practicability of our approaches, we apply them to the applications of function approximation and classification. For function approximation, we apply our approaches to model several nonlinear functions and real cases from measured input-output datasets. For classification, our approaches are applied to a problem of human object segmentation. A fuzzy self-clustering algorithm is used to divide the base frame of a video stream into a set of segments which are then categorized as foreground or background based on a combination of multiple criteria. Then, human objects in the base frame and the remaining frames of the video stream are precisely located by a fuzzy neural network which is constructed with the fuzzy rules previously obtained and is trained by our proposed HLA. Experimental results show that our approaches can improve the accuracy of human object identification in video streams and work well even when the human object presents no significant motion in an image sequence.

Submitted by Lucienne Costa (lucienneferreira@ufcg.edu.br) on 2018-09-13T20:25:53Z No. of bitstreams: 1 ALANA MARQUES DE MORAIS – TESE (PPGCC) 2018.pdf: 2316851 bytes, checksum: 52b3ebbbd6f1053b4978446f0a18d143 (MD5) / Made available in DSpace on 2018-09-13T20:25:53Z (GMT). No. of bitstreams: 1 ALANA MARQUES DE MORAIS – TESE (PPGCC) 2018.pdf: 2316851 bytes, checksum: 52b3ebbbd6f1053b4978446f0a18d143 (MD5) Previous issue date: 2018-04-20 / Capes / A evasão discente foi apontada, por grande parte das instituições brasileiras de Educação a Distância (EaD), como um dos maiores obstáculos enfrentados pelos cursos online da atualidade. Reflexo disto, foram as taxas de evasão elevadas, entre 25 e 50% em todos os tipos de curso online, em relação à educação presencial em 2015 no Brasil. Além disto, a problemática não se restringe ao contexto brasileiro, pois diversas organizações internacionais relataram índices tão alarmantes quando os nacionais. É possível afirmar, de acordo com a literatura, que as estratégias adotadas na minimização dos índices de desistência do discente muitas vezes se limitam à avaliação pontual de um recurso pedagógico (fórum, bate-papo, etc). Há ainda uma carência de estudos que discutam sobre as interações educacionais do aluno evadido e considerem as incertezas avaliativas envolvidas no processo. Diversas metodologias podem ser utilizadas pelo docente para lidar com a subjetividade e incerteza associadas ao processo avaliativo, tais como: lógica fuzzy, redes bayesianas, etc. O presente estudo destacou a lógica fuzzy (ou difusa), pois esta técnica lida com a ambiguidade encontrada na avaliação presente na EaD. Esta modelagem permite que estados indeterminados possam ser tratados por dispositivos de controle e conceitos não quantificáveis possam ser avaliados pelo especialista. Assim, o presente estudo propôs uma abordagem avaliativa (FuzzySD) voltada ao docente para monitorar as turmas virtuais na predição e combate à evasão discente em cursos online. Para tanto, a pesquisa se subsidiou nas informações relacionadas às ações dos alunos durante o processo de ensino-aprendizagem. A análise realizada envolve fatores comportamentais do estudante com base em múltiplos critérios. Estes critérios comportamentais foram fundamentais na descrição das três métricas principais ao entendimento da evasão do aluno no contexto estudado, a saber: Autorregulação, Interação e Motivação do aluno. A avaliação com a FuzzySD foi realizada por meio da análise de precisão em recortes da amostra com base em dados de cursos online do IFPB Virtual. A precisão da abordagem, analisando as amostras teste, proporciona valores entre 65% e 95% de acertos. Por fim, a partir dos resultados, foram discutidos os desafios da aplicação de um módulo avaliativo multidimensional diante do contexto de cursos online brasileiros. / The student’s dropout in online courses was pointed out as one of the biggest issue in the area according to the educational organizations. The main result was the high dropout rates in these courses compared to face-to-face courses in 2015 in Brazil (between 25% and 50%). Moreover, this challenge belongs to the virtual learning worldwide. In this sense, the literature review showed that several approaches are adopted to minimize the student’s dropout rates. However, they limited their focus on a learning unique resource (forum, chat, upload, download, submissions). Regarding these approaches, the current research detected an insufficiency of studies about the educational interactions of the dropout student and uncertainties in the assessment. Many methodologies are used by teachers to manage a subjectivity and uncertainty associated to the evaluation process, such as: fuzzy logic, Bayesian networks, Markov chains, etc. This thesis emphasized the fuzzy logic, because this technique solves the ambiguity found in the learning evaluation by the expert. We present an assessment framework (FuzzySD) to track the virtual classroom and to support the teacher to make a prediction of students’ dropout in online courses. In this sense, we analyzed the information related to the students’ actions during the teaching-learning process in online courses. It can be denominated as multidimensional analysis because we assessed multiple variables related to the student’s behavior. Such behavioral variables were fundamental in the definition of three main metrics to the understanding of student avoidance, they are: Self-regulation, Interaction and Motivation. It is important to consider a presence of fuzzy logic to deal with the input metrics of the FuzzySD framework. The evaluation of FuzzySD was performed by means of precision analysis on sample cut-outs based on dataset from courses of “IFPB Virtual”. The accuracy of the approach is analyzed as results of its own indicators between 65% and 95% of hits. Finally, the results discussed about the challenges of applying a multidimensional evaluative module to the context of Brazilian online courses.

This work describes a procedure for a soft sensor design to predict the top composition of a methanol-water distillation column. Soft sensor is a mathematical model that is used to estimate variables of interest from secondary variables easy to measure. This technique comes from an operational difficulty or high cost obtaining the desired variable. The approach to build a soft sensor was an artificial intelligence modeling, a black-box type, using a hybrid neuro-fuzzy technique. The data acquisition to train and validate the soft sensor comes from a mathematical model validated from pilot plat data. One of the limitations of neuro-fuzzy system is that it works with a limited number of inputs, depending on the combinatorial explosion of fuzzy rules. To minimize these effects and to reduce the number of rules in the training data sets of virtual analyzer, a data clustering technique called substractive clustering was used. To obtain a better performance of soft sensor for the dynamic process, distillation column, a regression of lone sampling time in selected variables was used, changing the number of entries from 9 to 18 variables, nine variables at actual sampling time and nine variables at previous sampling time. The distillation column is a good process for the present study because composition measurements are the main objective of this process and are difficult to obtain. The computational strategy for a soft sensor design produced good results in estimating the top composition of the methanol-water distillation column. / Este trabalho descreve um procedimento para o desenvolvimento de um analisador virtual, para predição da composição de topo de uma coluna destilação metanol-água em uma planta piloto. Analisador virtual é um modelo matemático que é usado para estimar variáveis de interesse a partir de variáveis secundárias de fácil medição. Esta tecnologia surge de uma real dificuldade operacional ou do alto custo de obtenção da variável desejada. O modelo utilizado nesta abordagem de construção do analisador virtual utiliza técnicas de sistemas inteligentes, tipo caixa preta, através da técnica híbrida neuro-fuzzy. A aquisição dos dados para treinar e validar o analisador virtual foi feita através de um modelo matemático validado a partir de dados experimentais da planta piloto. Uma das limitações do sistema neuro-fuzzy é que ele trabalha com um número limitado de entradas, dependendo da explosão combinatória das regras fuzzy. Para minimizar estes efeitos e conseguir reduzir o número de regras nos conjuntos de treinamento da rede neuro-fuzzy, foi utilizada a técnica de agrupamento de dados, denominada agrupamento substrativo. Com a intenção de se obter um melhor desempenho do analisador virtual no processo dinâmico, que é a coluna de destilação, foi empregada uma regressão de um tempo de amostragem nas variáveis de entrada selecionadas, alterando o número de entradas de 9 para 18, sendo 9 variáveis no tempo de amostragem atual e 9 variáveis em um tempo de amostragem anterior. O processo de destilação mostrou-se adequado para o presente estudo, pois as medições de composições são de difíceis obtenções. A estratégia computacional para um projeto de analisador virtual produziu bons resultados, de forma a estimar a composição do topo da coluna de destilação binária metanol-água.