Com o crescimento das empresas que fazem uso das tecnologias de bancos de dados, os administradores destes bancos de dados criam novos esquemas a cada instante, e na maioria dos casos não existe uma normalização ou procedimentos formais para que tal tarefa seja desempenhada de forma homogênea, resultando assim em bases de dados incompatíveis, o que dificulta a troca de dados entre as mesmas. Quando os Sistemas de Bancos de Dados (SBD) são projetados e implementados independentemente, é normal que existam incompatibilidades entre os dados de diferentes SBD. Como principais conflitos existentes nos esquemas de SBD, podem ser citados problemas relacionados aos nomes dos atributos, armazenamento em diferentes unidades de medida, diferentes níveis de detalhes, atributos diferentes com mesmo nome ou atributos iguais com nomes diferentes, tipos de dado diferentes, tamanho, precisão, etc. Estes problemas comprometem a qualidade da informação e geram maiores custos em relação à manutenção dos dados. Estes problemas são conseqüências de atributos especificados de forma redundante. Estes fatos têm provocado grande interesse em descobrir conhecimento em banco de dados para identificar informações semanticamente equivalentes armazenadas nos esquemas. O processo capaz de descobrir este conhecimento em banco de dados denomina-se DCDB (Descoberta de Conhecimento em Bancos de Dados). As ferramentas disponíveis para a execução das tarefas de DCDB são genéricas e derivadas de outras áreas do conhecimento, em especial, da estatística e inteligência artificial. As redes neurais artificiais (RNA) têm sido utilizadas em sistemas cujo propósito é a identificação de padrões, antes desconhecidos. Estas redes podem aprender similaridades entre os dados, diretamente de suas instâncias, sem conhecimento a priori. Uma RNA que tem sido usada com êxito para identificar equivalência semântica é o Mapa Auto-Organizável (SOM). Esta pesquisa objetiva descobrir, de modo semi-automatizado, equivalência semântica entre atributos de bases de dados, contribuindo para o gerenciamento e integração das mesmas. O resultado da pesquisa gerou uma sistemática para o processo de descoberta e uma ferramenta que a implementa. / With the increasing number of companies using database technologies, the database’s administrators create new schemes at every moment, and in most cases there are no normalization or formal procedures to do this task in a homogeneous form, it results in incompatible databases, that difficult data exchange. When the Database Systems (DBS) are projected and implemented independently, it is normal that data incompatibilities among different DBS. Problems related to the names of the attributes, storage in different measurement units, different levels of detail, different attributes with the same name or equal attributes with different names, different type of data, size, precision, etc, can be cited as main conflicts existing in the DBS schemes. These problems compromise the quality information and generate higher costs regarding the data maintenance. These problems arise as the consequence of redundant attributes’ specification. These facts have caused great interest in discovering knowledge in database to identify information semantically equivalent stored in schemes. The process capable to discover this knowledge in database is called KDD (Knowledge Discovery in Database). The available tools to do KDD tasks are generic and derived from other areas of knowledge, in special, statistics and artificial intelligence. The artificial neural networks (ANN) have been used in systems which aim is the identification of previously unknown patterns. These networks can learn similarities among the data directly from instances, without a priori knowledge. An ANN that has been used with success to identify semantic equivalence is the Self-Organizing Map (SOM). This research aims to discover, in a semi-automatic way, semantic equivalence on database attributes, contributing for the management and integration of these databases. This work resulted in a systematic for the discovery process and a tool that implements it.

Com o crescimento das empresas que fazem uso das tecnologias de bancos de dados, os administradores destes bancos de dados criam novos esquemas a cada instante, e na maioria dos casos não existe uma normalização ou procedimentos formais para que tal tarefa seja desempenhada de forma homogênea, resultando assim em bases de dados incompatíveis, o que dificulta a troca de dados entre as mesmas. Quando os Sistemas de Bancos de Dados (SBD) são projetados e implementados independentemente, é normal que existam incompatibilidades entre os dados de diferentes SBD. Como principais conflitos existentes nos esquemas de SBD, podem ser citados problemas relacionados aos nomes dos atributos, armazenamento em diferentes unidades de medida, diferentes níveis de detalhes, atributos diferentes com mesmo nome ou atributos iguais com nomes diferentes, tipos de dado diferentes, tamanho, precisão, etc. Estes problemas comprometem a qualidade da informação e geram maiores custos em relação à manutenção dos dados. Estes problemas são conseqüências de atributos especificados de forma redundante. Estes fatos têm provocado grande interesse em descobrir conhecimento em banco de dados para identificar informações semanticamente equivalentes armazenadas nos esquemas. O processo capaz de descobrir este conhecimento em banco de dados denomina-se DCDB (Descoberta de Conhecimento em Bancos de Dados). As ferramentas disponíveis para a execução das tarefas de DCDB são genéricas e derivadas de outras áreas do conhecimento, em especial, da estatística e inteligência artificial. As redes neurais artificiais (RNA) têm sido utilizadas em sistemas cujo propósito é a identificação de padrões, antes desconhecidos. Estas redes podem aprender similaridades entre os dados, diretamente de suas instâncias, sem conhecimento a priori. Uma RNA que tem sido usada com êxito para identificar equivalência semântica é o Mapa Auto-Organizável (SOM). Esta pesquisa objetiva descobrir, de modo semi-automatizado, equivalência semântica entre atributos de bases de dados, contribuindo para o gerenciamento e integração das mesmas. O resultado da pesquisa gerou uma sistemática para o processo de descoberta e uma ferramenta que a implementa. / With the increasing number of companies using database technologies, the database’s administrators create new schemes at every moment, and in most cases there are no normalization or formal procedures to do this task in a homogeneous form, it results in incompatible databases, that difficult data exchange. When the Database Systems (DBS) are projected and implemented independently, it is normal that data incompatibilities among different DBS. Problems related to the names of the attributes, storage in different measurement units, different levels of detail, different attributes with the same name or equal attributes with different names, different type of data, size, precision, etc, can be cited as main conflicts existing in the DBS schemes. These problems compromise the quality information and generate higher costs regarding the data maintenance. These problems arise as the consequence of redundant attributes’ specification. These facts have caused great interest in discovering knowledge in database to identify information semantically equivalent stored in schemes. The process capable to discover this knowledge in database is called KDD (Knowledge Discovery in Database). The available tools to do KDD tasks are generic and derived from other areas of knowledge, in special, statistics and artificial intelligence. The artificial neural networks (ANN) have been used in systems which aim is the identification of previously unknown patterns. These networks can learn similarities among the data directly from instances, without a priori knowledge. An ANN that has been used with success to identify semantic equivalence is the Self-Organizing Map (SOM). This research aims to discover, in a semi-automatic way, semantic equivalence on database attributes, contributing for the management and integration of these databases. This work resulted in a systematic for the discovery process and a tool that implements it.

Com o crescimento das empresas que fazem uso das tecnologias de bancos de dados, os administradores destes bancos de dados criam novos esquemas a cada instante, e na maioria dos casos não existe uma normalização ou procedimentos formais para que tal tarefa seja desempenhada de forma homogênea, resultando assim em bases de dados incompatíveis, o que dificulta a troca de dados entre as mesmas. Quando os Sistemas de Bancos de Dados (SBD) são projetados e implementados independentemente, é normal que existam incompatibilidades entre os dados de diferentes SBD. Como principais conflitos existentes nos esquemas de SBD, podem ser citados problemas relacionados aos nomes dos atributos, armazenamento em diferentes unidades de medida, diferentes níveis de detalhes, atributos diferentes com mesmo nome ou atributos iguais com nomes diferentes, tipos de dado diferentes, tamanho, precisão, etc. Estes problemas comprometem a qualidade da informação e geram maiores custos em relação à manutenção dos dados. Estes problemas são conseqüências de atributos especificados de forma redundante. Estes fatos têm provocado grande interesse em descobrir conhecimento em banco de dados para identificar informações semanticamente equivalentes armazenadas nos esquemas. O processo capaz de descobrir este conhecimento em banco de dados denomina-se DCDB (Descoberta de Conhecimento em Bancos de Dados). As ferramentas disponíveis para a execução das tarefas de DCDB são genéricas e derivadas de outras áreas do conhecimento, em especial, da estatística e inteligência artificial. As redes neurais artificiais (RNA) têm sido utilizadas em sistemas cujo propósito é a identificação de padrões, antes desconhecidos. Estas redes podem aprender similaridades entre os dados, diretamente de suas instâncias, sem conhecimento a priori. Uma RNA que tem sido usada com êxito para identificar equivalência semântica é o Mapa Auto-Organizável (SOM). Esta pesquisa objetiva descobrir, de modo semi-automatizado, equivalência semântica entre atributos de bases de dados, contribuindo para o gerenciamento e integração das mesmas. O resultado da pesquisa gerou uma sistemática para o processo de descoberta e uma ferramenta que a implementa. / With the increasing number of companies using database technologies, the database’s administrators create new schemes at every moment, and in most cases there are no normalization or formal procedures to do this task in a homogeneous form, it results in incompatible databases, that difficult data exchange. When the Database Systems (DBS) are projected and implemented independently, it is normal that data incompatibilities among different DBS. Problems related to the names of the attributes, storage in different measurement units, different levels of detail, different attributes with the same name or equal attributes with different names, different type of data, size, precision, etc, can be cited as main conflicts existing in the DBS schemes. These problems compromise the quality information and generate higher costs regarding the data maintenance. These problems arise as the consequence of redundant attributes’ specification. These facts have caused great interest in discovering knowledge in database to identify information semantically equivalent stored in schemes. The process capable to discover this knowledge in database is called KDD (Knowledge Discovery in Database). The available tools to do KDD tasks are generic and derived from other areas of knowledge, in special, statistics and artificial intelligence. The artificial neural networks (ANN) have been used in systems which aim is the identification of previously unknown patterns. These networks can learn similarities among the data directly from instances, without a priori knowledge. An ANN that has been used with success to identify semantic equivalence is the Self-Organizing Map (SOM). This research aims to discover, in a semi-automatic way, semantic equivalence on database attributes, contributing for the management and integration of these databases. This work resulted in a systematic for the discovery process and a tool that implements it.

Throughout last decades, knowledge discovery from databases as one of the information and communicaiton technologies' disciplines has developed into its current state being showed increasing interest not only by major business corporates. Presented diploma thesis deals with problematique of data mining while paying prime attention to its practical utilization within business environment. Thesis objective is to review possibilities of data mining applications and to decompose implementation techniques focusing on specific data mining methods and algorithms as well as adaptation of business processes. This objective is subject of theoretical part of thesis focusing on principles of data mining, knowledge discovery from databases process, data mining commonly used methods and algorithms and finally tasks typically implemented in this domain. Further objective consists in presenting data mining benefits on the model example that is being displayed in the practical part of the thesis. Besides created data mining models evalution, practical part contains also design of subsequent steps that would enable higher efficiency in some specific areas of given business. I believe previous point together with characterization of knowledge discovery in databases process to be considered as the most beneficial one's of the thesis.

This thesis applies the data mining method of decision rules on data from Consulting centre for Metabolism disorders from University hospital Pilsen. As a tool is used the system LISp-Miner, developed at University of Economics, Prague. Decision rules found are evaluated by a specialist. The main parts of this thesis are followings: an overview on main data mining methods and results evalutation methods, description of the data mining method application on data and description and evaluation of results.

Les travaux de recherche présentés dans ce mémoire s’inscrivent dans le cadre de la valorisation des connaissances issues des expériences passées afin d’améliorer les performances des processus industriels. La connaissance est considérée aujourd'hui comme une ressource stratégique importante pouvant apporter un avantage concurrentiel décisif aux organisations. La gestion des connaissances (et en particulier le retour d’expérience) permet de préserver et de valoriser des informations liées aux activités d’une entreprise afin d’aider la prise de décision et de créer de nouvelles connaissances à partir du patrimoine immatériel de l’organisation. Dans ce contexte, les progrès des technologies de l’information et de la communication jouent un rôle essentiel dans la collecte et la gestion des connaissances. L’implémentation généralisée des systèmes d’information industriels, tels que les ERP (Enterprise Resource Planning), rend en effet disponible un grand volume d’informations issues des événements ou des faits passés, dont la réutilisation devient un enjeu majeur. Toutefois, ces fragments de connaissances (les expériences passées) sont très contextualisés et nécessitent des méthodologies bien précises pour être généralisés. Etant donné le potentiel des informations recueillies dans les entreprises en tant que source de nouvelles connaissances, nous proposons dans ce travail une démarche originale permettant de générer de nouvelles connaissances tirées de l’analyse des expériences passées, en nous appuyant sur la complémentarité de deux courants scientifiques : la démarche de Retour d’Expérience (REx) et les techniques d’Extraction de Connaissances à partir de Données (ECD). Le couplage REx-ECD proposé porte principalement sur : i) la modélisation des expériences recueillies à l’aide d’un formalisme de représentation de connaissances afin de faciliter leur future exploitation, et ii) l’application de techniques relatives à la fouille de données (ou data mining) afin d’extraire des expériences de nouvelles connaissances sous la forme de règles. Ces règles doivent nécessairement être évaluées et validées par les experts du domaine avant leur réutilisation et/ou leur intégration dans le système industriel. Tout au long de cette démarche, nous avons donné une place privilégiée aux Graphes Conceptuels (GCs), formalisme de représentation des connaissances choisi pour faciliter le stockage, le traitement et la compréhension des connaissances extraites par l’utilisateur, en vue d’une exploitation future. Ce mémoire s’articule en quatre chapitres. Le premier constitue un état de l’art abordant les généralités des deux courants scientifiques qui contribuent à notre proposition : le REx et les techniques d’ECD. Le second chapitre présente la démarche REx-ECD proposée, ainsi que les outils mis en œuvre pour la génération de nouvelles connaissances afin de valoriser les informations disponibles décrivant les expériences passées. Le troisième chapitre présente une méthodologie structurée pour interpréter et évaluer l’intérêt des connaissances extraites lors de la phase de post-traitement du processus d’ECD. Finalement, le dernier chapitre expose des cas réels d’application de la démarche proposée à des interventions de maintenance industrielle. / The research work presented in this thesis relates to knowledge extraction from past experiences in order to improve the performance of industrial process. Knowledge is nowadays considered as an important strategic resource providing a decisive competitive advantage to organizations. Knowledge management (especially the experience feedback) is used to preserve and enhance the information related to a company’s activities in order to support decision-making and create new knowledge from the intangible heritage of the organization. In that context, advances in information and communication technologies play an essential role for gathering and processing knowledge. The generalised implementation of industrial information systems such as ERPs (Enterprise Resource Planning) make available a large amount of data related to past events or historical facts, which reuse is becoming a major issue. However, these fragments of knowledge (past experiences) are highly contextualized and require specific methodologies for being generalized. Taking into account the great potential of the information collected in companies as a source of new knowledge, we suggest in this work an original approach to generate new knowledge based on the analysis of past experiences, taking into account the complementarity of two scientific threads: Experience Feedback (EF) and Knowledge Discovery techniques from Databases (KDD). The suggested EF-KDD combination focuses mainly on: i) modelling the experiences collected using a knowledge representation formalism in order to facilitate their future exploitation, and ii) applying techniques related to data mining in order to extract new knowledge in the form of rules. These rules must necessarily be evaluated and validated by experts of the industrial domain before their reuse and/or integration into the industrial system. Throughout this approach, we have given a privileged position to Conceptual Graphs (CGs), knowledge representation formalism chosen in order to facilitate the storage, processing and understanding of the extracted knowledge by the user for future exploitation. This thesis is divided into four chapters. The first chapter is a state of the art addressing the generalities of the two scientific threads that contribute to our proposal: EF and KDD. The second chapter presents the EF-KDD suggested approach and the tools used for the generation of new knowledge, in order to exploit the available information describing past experiences. The third chapter suggests a structured methodology for interpreting and evaluating the usefulness of the extracted knowledge during the post-processing phase in the KDD process. Finally, the last chapter discusses real case studies dealing with the industrial maintenance domain, on which the proposed approach has been applied.

La recherche de règles d’association intéressantes est un domaine important et actif en fouille de données. Puisque les algorithmes utilisés en extraction de connaissances à partir de données (ECD), ont tendance à générer un nombre important de règles, il est difficile à l’utilisateur de sélectionner par lui même les connaissances réellement intéressantes. Pour répondre à ce problème, un post-filtrage automatique des règles s’avère essentiel pour réduire fortement leur nombre. D’où la proposition de nombreuses mesures d’intérêt dans la littérature, parmi lesquelles l’utilisateur est supposé choisir celle qui est la plus appropriée à ses objectifs. Comme l’intérêt dépend à la fois des préférences de l’utilisateur et des données, les mesures ont été répertoriées en deux catégories : les mesures subjectives (orientées utilisateur ) et les mesures objectives (orientées données). Nous nous focalisons sur l’étude des mesures objectives. Néanmoins, il existe une pléthore de mesures objectives dans la littérature, ce qui ne facilite pas le ou les choix de l’utilisateur. Ainsi, notre objectif est d’aider l’utilisateur, dans sa problématique de sélection de mesures objectives, par une approche par catégorisation. La thèse développe deux approches pour assister l’utilisateur dans sa problématique de choix de mesures objectives : (1) étude formelle suite à la définition d’un ensemble de propriétés de mesures qui conduisent à une bonne évaluation de celles-ci ; (2) étude expérimentale du comportement des différentes mesures d’intérêt à partir du point de vue d’analyse de données. Pour ce qui concerne la première approche, nous réalisons une étude théorique approfondie d’un grand nombre de mesures selon plusieurs propriétés formelles. Pour ce faire, nous proposons tout d’abord une formalisation de ces propriétés afin de lever toute ambiguïté sur celles-ci. Ensuite, nous étudions, pour différentes mesures d’intérêt objectives, la présence ou l’absence de propriétés caractéristiques appropriées. L’évaluation des mesures est alors un point de départ pour une catégorisation de celle-ci. Différentes méthodes de classification ont été appliquées : (i) méthodes sans recouvrement (CAH et k-moyennes) qui permettent l’obtention de groupes de mesures disjoints, (ii) méthode avec recouvrement (analyse factorielle booléenne) qui permet d’obtenir des groupes de mesures qui se chevauchent. Pour ce qui concerne la seconde approche, nous proposons une étude empirique du comportement d’une soixantaine de mesures sur des jeux de données de nature différente. Ainsi, nous proposons une méthodologie expérimentale, où nous cherchons à identifier les groupes de mesures qui possèdent, empiriquement, un comportement semblable. Nous effectuons par la suite une confrontation avec les deux résultats de classification, formel et empirique dans le but de valider et mettre en valeur notre première approche. Les deux approches sont complémentaires, dans l’optique d’aider l’utilisateur à effectuer le bon choix de la mesure d’intérêt adaptée à son application. / The search for interesting association rules is an important and active field in data mining. Since knowledge discovery from databases used algorithms (KDD) tend to generate a large number of rules, it is difficult for the user to select by himself the really interesting knowledge. To address this problem, an automatic post-filtering rules is essential to significantly reduce their number. Hence, many interestingness measures have been proposed in the literature in order to filter and/or sort discovered rules. As interestingness depends on both user preferences and data, interestingness measures were classified into two categories : subjective measures (user-driven) and objective measures (data-driven). We focus on the study of objective measures. Nevertheless, there are a plethora of objective measures in the literature, which increase the user’s difficulty for choosing the appropriate measure. Thus, our goal is to avoid such difficulty by proposing groups of similar measures by means of categorization approaches. The thesis presents two approaches to assist the user in his problematic of objective measures choice : (1) formal study as per the definition of a set of measures properties that lead to a good measure evaluation ; (2) experimental study of the behavior of various interestingness measures from data analysispoint of view. Regarding the first approach, we perform a thorough theoretical study of a large number of measures in several formal properties. To do this, we offer first of all a formalization of these properties in order to remove any ambiguity about them. We then study for various objective interestingness measures, the presence or absence of appropriate characteristic properties. Interestingness measures evaluation is therefore a starting point for measures categorization. Different clustering methods have been applied : (i) non overlapping methods (CAH and k-means) which allow to obtain disjoint groups of measures, (ii) overlapping method (Boolean factor analysis) that provides overlapping groups of measures. Regarding the second approach, we propose an empirical study of the behavior of about sixty measures on datasets with different nature. Thus, we propose an experimental methodology, from which we seek to identify groups of measures that have empirically similar behavior. We do next confrontation with the two classification results, formal and empirical in order to validate and enhance our first approach. Both approaches are complementary, in order to help the user making the right choice of the appropriate interestingness measure to his application.