Abstract : The research works presented here deal with the prediction and the performance improvement of parallel kinematic machine tools in order to produce machine parts with a specified quality level and in a minimum time. The problem treated allows determining the structure influence on the productivity and the machined part quality in the context of High Speed Machining for automotive and aeronautical parts.So, these works propose improvements along two fundamental ways : - modelling of the machine tool anisotropic behaviour ; - new methods of process adaptation. These approaches lead in a first time to a study of the time taken by the linking tool movement between cutting operations. The developed method is based on the definition of a kinematic model of linking tool movements, specific to each machine-tool and based on a command law defined in the joint workspace. A helpful resource for the setting up of machining with a parallel kinematic machine tool is then proposed. It is based on a numerical model of the kinematic behaviour using a command law of the movement defined in the programming workspace and providing a prediction of machining time. The process optimization is also based on the machining quality prediction brought by an experimental model enhanced by a measurement campaign realized on the considered machine tool. These approaches are applied to industrial parts with the PCI Tripteor X7 machine-tool. Their originality lies in the improvement of parallel kinematic machines tool performances from an analysis of the machine behaviour during the machining, and consequently allows extending their application field.