Les entrepôts de données (ED) sont devenus une technologie mature. L'accentuation des demandes d'analyse est motivée par l'évolution technologique, Les nouveaux paradigmes de programmation et L'ingénierie Dirigée par les Modèles (!DM). Avant d'utiliser ces progrès technologiques, l'entrepôt de données doit être construit et préparé pour sa bonne exploitation.La phase de construction a vu l'utilisation massive des efforts de description et de méta-modélisation afin de faciliter la définition des correspondances entre les schémas locaux des sources de données et le schéma de l'ED et de réduire l'hétérogénéité entre les sources. La phase d'exploitation et sa tâche physique, en particulier n'ont pas eu la même utilisation des solutions de description et de méta-modélisation, bien qu'elle est considérée comme un tunnel de toutes les phases de cycle de vie de conception d,un ED. Durant cette phase; des modèles de coût mathématiques sont utilisés pour quantifier la qualité des solutions proposées. Le développement de ces derniers nécessite des efforts de collection et d'analyse des paramètres pertinents.Pour bien simuler le fonctionnement d'un ED, toutes les dimensions d'un SGBD doivent être intégrées. Dans cette thèse, nous proposons de décrire en détail ces dimensions avec des mécanismes de méta-modélisation. Vu la similarité et la hiérarchisation gui existent entre les supports de stockage, nous avons développé une ontologie de domaine dédiée aux supports de stockage.Elle permet d'expliciter leurs propriétés. Les similarités entre ces supports nous a motivé à hybrider le cache mémoire avec les mémoires flashs pour augmenter sa capacité afin de stocker un nombre important de résultats intermédiaires partagés par plusieurs requêtes décisionnelles. La réutilisation de ces résultats permet d'augmenter la performance du SGBD. Nos contributions sont validées à l'aide des expérimentations en utilisant nos modèles de coût théoriques et le SGBD Oracle. / Data warehouses (DW) have become a mature technology. The emphasis of the analysis requests is driven by technological change, the new programmig paradigms and ModelDriven Engineering (MDI). Before using these technological advances, the DW must be buil tand prepared for its proper operation. The construction phase bas seen massive description efforts and meta modeling to facilitate the definition of correspondence between local data sources schemas and DW schema and to reduce heterogeneity between sources. Despite its importance in all stages of the design life cycle of an DW, the operational phase and in particular its physical task, did not have the same interest in term of description and meta modeling. During this phase, mathematical cost models are used to quantify the quality of the solutions proposed. The development of these models requires collection efforts and analysis of relevant parameters. To simulate the operation of a DW, all the dimensions of a DBMS must be integrated. In this thesis, we propose to describe in detail these dimensions with meta-modeling mechanisms. Given the singularity and hierarchy between storage media, we have developed an ontology dedicated to storage media, which makes explicit their properties. The similarities between these supports motivated us to develop a hybrid cache based on flash memory. This increases the cache ability to store a large number of intermediate results shared by multiple decision-support queries. The reuse of these results will increase the overall performance of fue DBMS. Our contributions are validated with experiments using our theoretical cost models and the Oracle DBMS.

Les langages dédiés (DSL) sont de plus en plus utilisés parce qu’ils permettent aux utilisateurs qui ne sont pas des experts en programmation d’exprimer des solutions avec des langages simples qui capturent l’expertise de leur domaine. C’est encore plus vrai pour les langages dédiés graphiques (DSML) qui ont un niveau d’abstraction plus élevé que les langages dédiés de programmation. Implémenter un DSML revient généralement à fournir un éditeur dédié qui permette aux utilisateurs de manipuler les abstractions de leur domaine (d’instancier le langage). Les expériences ont montré que l’implémentation d’un tel éditeur dédié graphique est coûteuse en termes de temps et de ressources humaines. Nous constatons que la plupart des plates-formes permettant de construire ce type d’éditeur (EMF/GMF, DSL Tools, Obeo Designer, ...) sont génériques. Elles essayent d’adresser le maximum de domaines possibles, ce qui les rend complexes et inadaptées à des cas d’utilisation spécifiques. Si la spécialisation aux domaines a été un succès pour les langages, pourquoi ne pas l’appliquer aux plates-formes de construction d’éditeurs ? Cela reviendrait à concevoir pour un domaine donné, une plate-forme permettant de construire facilement des éditeurs dédiés pour ce domaine. Cette plate-forme n’aurait pas les défauts d’une plate-forme totalement générique parce qu’elle serait restreinte au domaine ciblé. Ce type de plate-forme spécifique à un domaine, nous l’appelons Domain Specific Modeling Framework (DSMF). Le principal inconvénient d’un DSMF est qu’on ne peut l’utiliser que dans le cadre du domaine pour lequel il a été conçu. Cela implique qu’il faille construire un DSMF par domaine et c’est une solution coûteuse. Toutefois, nous pensons que cette approche sur les DSMF peut être généralisée afin d’adresser un grand nombre de domaines. Cette thèse a donc consisté à concevoir et à implanter un environnement qui permet de construire des DSMF de façon modulaire. / Domain Specific Languages (DSLs) are increasingly used in many fields as they allow users to express strategies without being programming experts. This is particularly true for graphical DSLs called Domain Specific Modeling Languages (DSMLs) which are more intuitive than programming DSLs. Implementing a DSML means providing a specific editor which allow users to express the language’s constructions (instantiate the language). Many experiments showed that implementing specific graphical editors is much manpower consuming. Our analysis is that most frameworks for building such editors (e.g. EMF/GMF) are generic, i.e. aim at fulfilling the requirements of any field, which leads to increased complexity and costs a lot in terms of development time. If domain specialization was successful for languages, why don’t we apply it to frameworks ? Specializing such a framework according to the constraints of a domain would allow keeping the definition of a specific editor simple, while fulfilling the requirements of the considered domain. Domain specific frameworks for building DSML editors in specific application fields is a promising approach. Such a framework does not have the limits of generic frameworks because it is restricted to a particular domain. It is more intuitive and simpler to use as it only proposes abstraction of the domain for building DSMLs. We call this type of framework Domain Specific Modeling Framework (DSMF). For example, if we consider the component domain, there are several DSMLs in this domain which share the same layout requirements. We implemented a DSMF for this family of DSMLs. This DSMF is specialized according to the constraints and layout requirements of the component domain (Components, connectors, Bindings, . . .). This specialization allows simple and rapid generation of specific editors devoted to component-based architectures. The principal drawback of a DSMF is its restricted scope to one specific domain. This approach requires to develop one DSMF per domain and the development cost can be significant. A solution may be to generalise the DSMF approach in order to address many application fields. We designed a Generic framework for building DSMFs in a modular way. This thesis is based on the implementation of this framework.

<p>Modeling is a complex process which is quite hard to do in a structured and controlled way. Many companies provide a set of guidelines for model structure, naming conventions and other modeling rules. Using meta-models to describe these guidelines makes it possible to check whether an UML model follows the guidelines or not. Providing this error checking of UML models is only one step on the way to making modeling software an even more valuable and powerful tool.</p><p>Moreover, by providing correction suggestions and automatic correction of these errors, we try to give the modeler as much help as possible in creating correct UML models. Since the area of model correction based on meta-models has not been researched earlier, we have taken an explorative approach. </p><p>The aim of the project is to create an extension of the program MetaModelAgent, by Objektfabriken, which is a meta-modeling plug-in for IBM Rational Software Architect. The thesis shows that error correction of UML models based on meta-models is a possible way to provide automatic checking of modeling guidelines. The developed prototype is able to give correction suggestions and automatic correction for many types of errors that can occur in a model.</p><p>The results imply that meta-model guided error correction techniques should be further researched and developed to enhance the functionality of existing modeling software.</p>

Thesis (MSc)--Stellenbosch University, 2014. / ENGLISH ABSTRACT: This thesis involved the development and application of network approaches for the construction, analysis and visualization of phylogenomic and transcriptomic networks. A co-evolutionary network model of grapevine genes was constructed based on three mechanisms of evolution. The investigation of local neighbourhoods of this network revealed groups of functionally related genes, illustrating that the multi-mechanism evolutionary model was identifying groups of potentially co-evolving genes. An extended network definition, namely 3-way networks, was investigated, in which edges model relationships between triplets of objects. Strategies for weighting and pruning these 3-way networks were developed and applied to a phylogenomic dataset of 211 bacterial genomes. These 3-way bacterial networks were compared to standard 2-way network models constructed from the same dataset. The 3-way networks modelled more complex relationships and revealed relationships which were missed by the two-way network models. Network meta-modelling was explored in which global network and node-bynode network comparison techniques were applied in order to investigate the effect of the similarity metric chosen on the topology of multiple types of networks, including transcriptomic and phylogenomic networks. Two new network comparison techniques were developed, namely PCA of Topology Profiles and Cross-Network Topological Overlap. PCA of Topology Profiles compares networks based on a selection of network topology indices, whereas Cross- Network Topological Overlap compares two networks on a node-by-node level, identifying nodes in two networks with similar neighbourhood topology and thus highlighting areas of the networks with conflicting topologies. These network comparison methods clearly indicated how the similarity metric chosen to weight the edges of the network influences the resulting network topology, consequently influencing the biological interpretation of the networks. / AFRIKAANSE OPSOMMING: Hierdie tesis hou verband met die ontwikkeling en toepassing van netwerk benaderings vir die konstruksie, analise en visualisering van filogenomiese en transkriptomiese netwerke. 'n Mede-evolusionêre netwerk model van wingerdstok gene is gebou, gebaseerd op drie meganismes van evolusie. Die ondersoek van plaaslike omgewings van die netwerk het groepe funksioneel verwante gene aan die lig gebring, wat daarop dui dat die multi-meganisme evolusionêre model groepe van potensieele mede-evolusieerende gene identifiseer. 'n Uitgebreide netwerk definisie, naamliks 3-gang netwerke, is ondersoek, waarin lyne die verhoudings tussen drieling voorwerpe voorstel. Strategieë vir weeg en snoei van hierdie 3-gang netwerke was ontwikkel en op 'n filogenomiese datastel van 211 bakteriële genome toegepas. Hierdie 3-gang bakteriële netwerke is met die standaard 2-gang netwerk modelle wat saamgestel is uit dieselfde datastel vergelyk. Die 3-gang netwerke het meer komplekse verhoudings gemodelleer en het verhoudings openbaar wat deur die tweerigting-netwerk modelle gemis is. Verder is netwerk meta-modellering ondersoek waarby globalle netwerk en punt-vir-punt netwerk vergelykings tegnieke toegepas is, met die doel om die effek van die ooreenkoms-maatstaf wat gekies is op die topologie van verskeie tipes netwerke, insluitend transcriptomic en filogenomiese netwerke, te bepaal. Twee nuwe netwerk-vergelyking tegnieke is ontwikkel, naamlik "PCA of Topology Profiles" en"Cross-Network Topological Overlap". PCA van Topologie Profiele vergelyk netwerke gebaseer op 'n seleksie van netwerk topologie indekse, terwyl Cross-netwerk Topologiese Oorvleuel vergelyk twee netwerke op 'n punt-vir-punt vlak, en identifiseer punte in twee netwerke met soortgelyke lokale topologie en dus lê klem op gebiede van die netwerke met botsende topologieë. Hierdie netwerk-vergelyking metodes dui duidelik aan hoe die ooreenkoms maatstaf wat gekies is om die lyne van die netwerk gewig te gee, die gevolglike netwerk topologie beïnvloed, wat weer die biologiese interpretasie van die netwerke kan beïnvloed.

Nous avons défini et réalisé avec l'approche MuVarch un environnement de (méta-)modélisation orientée vers la représentation multi-vues des architectures embarquées hétérogènes (de type "smartphone" par exemple). En plus de la vue architecturale de base, support de toutes les autres, on considère les vues "performance", "consommation", "température", ainsi que la vue fonctionnelle "applicative" pour fournir des scénarios comportementaux de fonctionnement de la plate-forme. Il était important de savoir décrire en MuVarch comment les vues se raccrochent à la vue de base architecturale, et comment elle se relient également entre elles (relation entre consommation énergétique et température par exemple). L'objectif ultime est d'utiliser ce framework multi-vues et les différentes informations apportées par chacune, pour savoir supporter des politiques alternatives de mapping/allocation des tâches applicatives sur les ressources de l'architecture (la définition de ces politiques restant extérieure à nos travaux de thèse). La représentation adéquate de cette relation d'allocation forme donc un des aspects importants de nos travaux. / We introduced and realized with our MuVarch approach an heterogeneous (meta)modeling environment for multi-view representation of heterogeneous embedded architectures (of "smartphone" type for instance). In addition to the backbone architectural view supporting others, we considered performance, power, and thermal view. We introduced also the functional applicative view, to provide typical use cases for the architecture. It was important to describe in MuVarch our various views would connect to the basic one, and how they would mutually relate together as well (how temperature depends on power consumption for instance). The global objective was to let the framework consider alternative mapping/allocation strategies for applicative tasks on architectural resources (although the definition of such strategies themselves was out of the scope). The appropriate form of such an allocation relation, which may be quite involved, was thus an important aspect of this thesis.

Modeling is a complex process which is quite hard to do in a structured and controlled way. Many companies provide a set of guidelines for model structure, naming conventions and other modeling rules. Using meta-models to describe these guidelines makes it possible to check whether an UML model follows the guidelines or not. Providing this error checking of UML models is only one step on the way to making modeling software an even more valuable and powerful tool. Moreover, by providing correction suggestions and automatic correction of these errors, we try to give the modeler as much help as possible in creating correct UML models. Since the area of model correction based on meta-models has not been researched earlier, we have taken an explorative approach. The aim of the project is to create an extension of the program MetaModelAgent, by Objektfabriken, which is a meta-modeling plug-in for IBM Rational Software Architect. The thesis shows that error correction of UML models based on meta-models is a possible way to provide automatic checking of modeling guidelines. The developed prototype is able to give correction suggestions and automatic correction for many types of errors that can occur in a model. The results imply that meta-model guided error correction techniques should be further researched and developed to enhance the functionality of existing modeling software.

Enterprise models are usually developed with ambition to capture the current or desired situation in enterprises with respect to performed or planned processes, organizational structure (including organization units, roles and competences), products or services produced and IT systems available in the enterprise.The above aspects are mutually reflective. Such enterprise models are often represented in formal modeling languages, like UEML (Unified Enterprise Modeling Language) or GEM (General Entity Manipulator) language allowing for the development of applications, which interprets or compute them. Enterprise ontologies basically allow the representation of the same aspects of an enterprise (processes, organizational structure, products and systems). However, enterprise ontologies use another representation (like OWL- Web Ontology Language) and often are developed for other application purposes than enterprise model. The objective of this thesis is to develop strategies for transforming enterprise models into enterprise ontologies.  There should be maximum preservation of semantics and minimum loss of information during the process of transformation. On the basis of meta-model (model to model) transformation, we propose three elements mapping approaches. Each approach has a number of elements mapping rules. After comparative study the best suitable approach according to objective of this thesis is selected for implementation purpose. From a technical perspective, a tool named “EM2EO” is developed, which accepts an enterprise model as input and produces ontology as output.

Die Arbeit adressiert insgesamt drei Forschungsschwerpunkte. Der erste Schwerpunkt setzt sich mit zu entwickelnden BPMN-Erweiterungen auseinander und stellt deren methodische Implikationen im Rahmen der bestehenden Sprachstandards dar. Dies umfasst zum einen ganz konkrete Spracherweiterungen wie z. B. BPMN4CP, eine BPMN-Erweiterung zur multi-perspektivischen Modellierung von klinischen Behandlungspfaden. Zum anderen betrifft dieser Teil auch modellierungsmethodische Konsequenzen, um parallel sowohl die zugrunde liegende Sprache (d. h. das BPMN-Metamodell) als auch die Methode zur Erweiterungsentwicklung zu verbessern und somit den festgestellten Unzulänglichkeiten zu begegnen. Der zweite Schwerpunkt adressiert die Untersuchung von sprachunabhängigen Fragen der Erweiterbarkeit, welche sich entweder während der Bearbeitung des ersten Teils ergeben haben oder aus dessen Ergebnissen induktiv geschlossen wurden. Der Forschungsschwerpunkt fokussiert dabei insbesondere eine Konsolidierung bestehender Terminologien, die Beschreibung generisch anwendbarer Erweiterungsmechanismen sowie die nutzerorientierte Analyse eines potentiellen Erweiterungsbedarfs. Dieser Teil bereitet somit die Entwicklung einer generischen Erweiterungsmethode grundlegend vor. Hierzu zählt auch die fundamentale Auseinandersetzung mit Unternehmensmodellierungssprachen generell, da nur eine ganzheitliche, widerspruchsfreie und integrierte Sprachdefinition Erweiterungen überhaupt ermöglichen und gelingen lassen kann. Dies betrifft beispielsweise die Spezifikation der intendierten Semantik einer Sprache.

La séparation des préoccupations permet de considérer les sous-systèmes en isolation tout en conservant une vision globale du système. La variabilité des domaines impliqués implique donc (i) un effort de composition des méta-modèles hétérogènes représentant ces domaines, (ii) une gestion de la cohérence inter-domaine des modèles produits en isolation et (iii) une gestion de la multiplicité des cibles atteignables dans l'espace des solutions de chacun des domaines. Pour relever ces défis, nous présentons dans cette thèse une approche couvrant trois contributions appliquées sur le cas de la visualisation de données :- une approche de composition respectant l'isolation des domaines en tirant profit des méthodes d'intégration des Architectures Orientées Services (SOA). Les méta-modèles sont encapsulés dans des services, exposant le comportement pertinent via une interface à destination des experts du domaine ;- un moteur de règles métiers qui assure la gestion des interactions entre domaines et permet de détecter les incohérences inter-domaine et de faire remonter aux experts les informations nécessaires à leur résolution ;- une modélisation de la variabilité des produits par caractérisation qui permet de concrétiser les sous-systèmes vers des artefacts concrets.Pour valider l'approche, nous quantifions (i) le surcoût de l'encapsulation en service des domaines, (ii) l'impact de l'externalisation des interactions entre domaines, (iii) l'effort que les experts et l'intégrateur doivent fournir. Enfin, une expérience utilisateur mesure le gain lors de la concrétisation du système et son impact sur la satisfaction vis à vis des visualisations résultantes. / The growing complexity of software engineering leads to the use of separation of concerns, i.e. enable to consider manageable sub-systems while keeping an overview of the whole system. The domain variability involved in these system design imply : (i) to compose multiple heterogeneous metamodels dedicated to each domain, (ii) to handle cross-domain consistency of the model produced in isolation, (iii) and to tame the multiplicity of concrete artefact available in the solution space of each domain. To adress these challenges, we offer in this thesis an approach encompassing three contributions :- an isolation-compliant composition which benefits from Service Oriented Architecture (SOA) integration. Each domain metamodel is embedded in a service exposing the rele- vant behavior through an interface designed and used by domain experts ; - a business rule engine handling the interaction between domains and detecting cross-domain incon- sistency to provide relevant feedback to resolve it ;- a feature-based characterization of the products variability allowing to concretize each sub-system toward concrete artifacts. The contributions are applied on the sensor data visualization use case.We validate (i) the need for domain isolation preservation on a dashboard design project, then we quantify (ii) the overhead of the service encapsulation, (iii) the impact of the externalization of domain interactions, (iv) the effort required from the experts and the integrator. Finally, we proceed to a user experiment to measure the gain during the concretization of a sub-system, and the impact on the user satisfaction on the resulting visualisation widgets.

Engenharia Dirigida por Modelos, ou Model Driven Engineering (MDE), é uma abordagem para desenvolvimento de software a partir de modelos. Código-fonte ou artefatos executáveis são gerados de forma automática, total ou parcialmente, a partir de transformações descritas por funções de mapeamento. Com isso, obtém-se as vantagens de desenvolver software em nível de abstração maior em relação às linguagens de programação tradicionais e da possibilidade de gerar implementações do mesmo sistema para diversas plataformas a partir do mesmo modelo. Uma das áreas de pesquisa da MDE é a formalização de teorias e conceitos relacionados a essa abordagem, tais como modelos, metamodelos, relação de conformidade, operações sobre metamodelos, etc. Embora existam na literatura propostas de formalização, observam-se lacunas e falta de consenso geral, o que leva autores a introduzir suas próprias definições quando desejam apresentar desenvolvimentos ou deduções, as quais nem sempre se encaixam entre si. Alguns autores consideram que, enquanto uma completa formalização de conceitos e relações da MDE não for estabelecida, seus potenciais podem não ser plenamente atingidos. A partir de estudos de proposições existentes na literatura, identificação de lacunas e necessidades, este trabalho apresenta uma proposta de arcabouço teórico para MDE, baseada nas teorias de conjuntos e linguagens, e em lógica de primeira ordem. Conceitos e operações relacionados à MDE são definidos sobre uma base comum e, a partir dela, algumas formulações originais são desenvolvidas, tais como a função de mesclagem de metamodelos e uma definição de modelos executáveis, que podem servir como base para a construção de motores de execução de modelos, consolidando e ampliando o conhecimento da área teórica da MDE. Do ponto de vista de aplicabilidade prática, a proposta é validada por meio de uma ferramenta de prova de conceito, criada também como parte desta pesquisa, e pela apresentação de exemplos de aplicações. / Model Driven Engineering (MDE) is a software development approach in which models are essential artifacts to build software systems. Source code or executable artifacts are automatically generated, completely or partially, by transformations described by mapping functions. Two main advantages can be obtained: software development in a higher level of abstraction than that of traditional programming languages and the possibility of generating implementations of the same system for multiple platforms from the same source model. One of the research fields of MDE is the formalization of theories and concepts related to this approach, such as models, metamodels, conformity relationship, operations over metamodels, etc. Although there are proposals of formalization in the literature, some gaps and a lack of general consensus can be identified, which leads some authors to introduce their own definitions when they want to present their work. These definitions not always fit to each other. Some authors claim that unless a complete formalization of MDE concepts and relations is given, the potentials of this approach may not be fully unfolded. Starting from studies on existing proposals, gaps and requirements, this work proposes a theoretical framework for MDE based on the set theory, language theory and first order logic. Concepts and operations related to MDE are defined over a common basis and some original formulations are developed, such as the metamodel merging function and definitions about executable models, which can be used as a foundation to build model execution engines, consolidating and expanding the theoretical field of MDE. From the applicability point of view, the proposal is validated by a proof-of-concept tool, created as part of this research, and by examples of applications.