Cadre de conception multi-perspectives pour la modelisation des processus de genie logiciel

Cost and schedule of software projects are often overrun. This fact is closely-related to the software process and less to the final product as acknowledged by researchers and practitioners (Boehm, 1991; Barros et al., 2004; Roy, 2004; McConnell, 2009). However, it is also acknowledged that the software production is human and knowledge-intensive process, which means the nature of uncertainty. In order to build efficient process modeling tools, there is a need for conceptualization of process models to address different concerns such as a breakdown structure of activities, knowledge, simulation, alignment with CMMI practices and so on. Addressing those concerns provide a context for analysis, assessment and enhancement of process models, which in turn enhance the quality of software projects.;The first perspective tackles the issue of knowledge management integration within the process modeling as a mean to improve software development process and to mitigate risks related to the lack of knowledge. In fact, software processes are often termed as knowledge-intensive. Despite this, knowledge concepts are rarely taken into account in software process modeling. We propose an approach which takes into account the conceptual knowledge required to perform tasks. The approach adds knowledge attributes to existing relationships between roles, tasks, and artifacts. Comparison between the attributes of a given task provides information on the knowledge discrepancies gaps between the SPEM elements involved. This information could be used on practical level to support team management to make informed decisions.;The second perspective tackles the issue of process simulation. Indeed, planning of software project and its deliverables is closely-related to the software process. However SPEM-based software process modeling describes only structural concerns and does not support simulation.;We present an extension of a SPEM-based software process modeling with stochastic simulation aiming at: 1) Assessing risk by using Monte-Carlo simulation based on the same descriptive process model; 2) Reducing the gap between three communities: process modeling, process simulation and project management and 3) Allowing "What if" experiments to help the prediction of changes' impact on project schedule. The proposed approach considers duration of tasks as a random variable with certain probability distribution. The results outline the critical path of the process model and sensitivity analysis based on spearman's rank correlation between durations of tasks and the whole process.;One attempt to enhance the process modeling flexibility has been made with the tool called DSL4SPM (Domain-Specific Language for Software Process Modeling). DSL4SPM was developed and applied to bring order and structure to the software development process. It implements the Software & Systems Process Engineering Meta-model (SPEM 2.0) specification, and is characterized by: 1) a conceptual framework for designing processes in an abstract way; and 2) multi-view-oriented process modeling, which acknowledges the relevance of a multitude of issues in a process model. The conceptual framework emphasizes the relationships among the SPEM elements.;The third perspective introduces the alignment of the process models with the CMMI best practices using DSL4SPM tool. Furthermore, we used our tool to lead an exploratory study in order to understand how the users of software process perceive the intention of each task of the process according to CMMI components. The results of the study indicate that the users are able to align tasks with the CMMI process area, but there is great variability in alignment with the finest specific practices of CMMI. However, there is a need to clarify the intention of each task according to CMMI components to benefit from software process improvement (SPI) initiatives. Therefore, the integration of CMMI in the software process modeling offers a good framework for this purpose.;We believe that the framework and results presented in this thesis will provide a practical response to the needs of research in the field of software process modeling. The central idea is the conceptualization of process models from several perspectives. Such a conceptual approach with the three new perspectives, hopefully, will allow the design of robust, comprehensive and flexible processes, contributing therefore to the improvement of software engineering. The DSL4SPM tool that is the proof concept usability of the proposed approach will open new directions to software process modeling.