Avis de Soutenance de Monsieur Karn YONGSIRIWIT
informatique

dirigés par Monsieur Walid GAALOUL

Quand: Soutenance prévue le lundi 23 janvier 2017 à 14h00
Lieu : Télécom SudParis 9 rue Charles Fourier 91011 Evry Cedex, France
salle C06

Composition du jury proposé

Mots-clés :Modèle de processus,fouille du processus,allocation des ressources Cloud,fragment du processus métier,ontologie,base de connaissances

Résumé :
More and more organizations are adopting cloud-based Process-Aware Information Systems (PAIS) to manage and execute processes in the cloud as an environment to optimally share and deploy their applications. This is especially true for large organizations having branches operating in different regions with a considerable amount of similar processes. Such organizations need to support many variants of the same process due to their branches’ local culture, regulations, etc. However, developing new process variant from scratch is error-prone and time consuming. Motivated by the  »Design by Reuse » paradigm, branches may collaborate to develop new process variants by learning from their similar processes. These processes are often heterogeneous which prevent an easy and dynamic interoperability between different branches. A process variant is an adjustment of a process model in order to flexibly adapt to specific needs. Many researches in both academics and industry are aiming to facilitate the design of process variants. Several approaches have been developed to assist process designers by searching for similar business process models or using reference models. However, these approaches are cumbersome, time-consuming and error-prone. Likewise, such approaches recommend entire process models which are not handy for process designers who need to adjust a specific part of a process model. In fact, process designers can better develop process variants having an approach that recommends a well-selected set of activities from a process model, referred to as process fragment. Large organizations with multiple branches execute BP variants in the cloud as environment to optimally deploy and share common resources. However, these cloud resources may be described using different cloud resources description standards which prevent the interoperability between different branches. In this thesis, we address the above shortcomings by proposing an ontology-based approach to semantically populate a common knowledge base of processes and cloud resources and thus enable interoperability between organization’s branches. We construct our knowledge base built by extending existing ontologies. We thereafter propose an approach to mine such knowledge base to assist the development of BP variants. Furthermore, we adopt a genetic algorithm to optimally allocate cloud resources to BPs. To validate our approach, we develop two proof of concepts and perform experiments on real datasets. Experimental results show that our approach is feasible and accurate in real use-cases.