In recent decades, enterprises' information systems become more and more flooded by all kind of data: structured (databases, data warehouse), semi-structured (XML, server log files), and unstructured data (raw text, multimedia data). This has created new challenges for companies and for the scientific community. Including, how to understand and analyze such a mass of data to extract knowledge. Moreover, in an organization, a data mining project is usually conducted by several experts (domain experts, KDD experts, data experts...) who consequently manipulate several types of knowledge and know-how. They will have different objectives and preferences, different competences, and different visions of analyzed data and of KDD methods. Our objective in this thesis is to facilitate the KDD analyst task, and to improve coordination and comprehensibility between the different actors in a multi-view analysis as well as the reuse of KDD process in terms of viewpoints. Therefore, we propose a definition that makes explicit the notion of viewpoint in KDD and includes domain knowledge (analyzed domain and analyst domain) and context of analysis. Based on this definition, we propose the development of a set of semantic models that are structured in a Conceptual Model and allowing knowledge representation and management during a multi-view analysis. Our approach is based on a multi-criteria characterization of viewpoint in KDD. A characterization that is primarily designed to capture the objectives and context of analysis of the expert, guide the construction and execution of the KDD process, and then keep the trace, in the form of annotations, of reasoning made during a collaborative work. These annotations can be shared, compared and reused based on a set of semantic relations between viewpoints.
