Describir: A Hybrid Framework for the Sensitivity Analysis of Software-Defined Networking Performance Metrics Using Design of Experiments and Machine Learning Techniques