Technologies such as IoT and Big data use advanced representation models and methods to lead to coherent systems and softwares. Nevertheless, there is a substantial lack of approaches able to support uncertain data and fuzzy environment to build a bridge between physical objects, semantic real-world view and the systemic view. In this work, we propose a semantic driven approach to support a decisional system based on semantic representation and an accurate fuzzy reasoning using the Choquet Integral method. The proposed generic architecture takes into consideration the formal representation of the real world, users' needs and processes. Further, we focus on the relation between them and the dependence between different criteria. A detailed case study in the agriculture domain is also presented to showcase the interest of our proposal.