Flexible querying of DB allows to extend DBMS in order to support imprecision and flexibility in queries. Flexible queries use vague and imprecise terms which have been defined as fuzzy sets. However, there is no consensus on memberships functions generation. Most of the proposed methods require expert intervention. This thesis is devised in two parts. In the first part, we propose a clustering based approach for automatic and incremental membership functions generation. We have proposed the clustering method CLUSTERDB* which evaluates clustering quality underway clusters generation. Moreover, we propose incremental updates of partitions and membership functions after insertion or deletion of a new object. The second part of this thesis uses these functions and Formal Concepts Analysis in flexible and cooperative querying. In case of empty answers, we formally detect the failure reasons and we generate approximative queries with their answers. These queries help the user to formulate new queries having answers. The different proposed approaches are implemented and experimented with several databases. The experimentation results are encouraging.