The problem of configuring the variability models is pervasive in plenty of domains. Renault, a leading automobile manufacturer, has developed an internal product configuration system to model its vehicle diversity. This system is based on the well-known knowledge compilation approach and is associated with a set of parameters. Different input parameters have a strong influence on the system's performance. The parameters actually used are determined manually. Our work aims to study and determine these parameters automatically. This paper studies Renault's variability models and product configuration system and presents a parameter prediction model for this system. The results show the predicted parameters' competitiveness compared with the parameters by default