Software Product Lines (SPL) aim at deriving software architectures or systems from a software artifact base. Configuring the SPL to derive a new product is now usually done by selecting appropriate software features in a kind of models, called feature models. In some situations, a feature represents a software artifact associated to an element e of a context the software product will manage. Such a feature and its associated software artifact may be cloned according to the number of occurrences of e in the context and constraints have to be respected. Hence, the feature model proposed to users for configuration has to be adapted in a new dedicated phase according to the context elements. We propose a model-driven engineering approach for transforming a generic feature model according to a context model that a derived software product will manage. More precisely this paper describes an original model transformation able to generate context specific feature models including duplicated features, and removing inappropriate features. Our transformation is validated on a smart building optimization software case study.