In this presentation, we will introduce a novel data-driven adaptive robust optimization framework that organically integrate machine learning techniques and optimization under uncertainty. A Bayesian nonparametric model – the Dirichlet process mixture model – is adopted and combined with a variational inference algorithm to extract the information embedded within uncertainty data. We further propose a data-driven approach for defining uncertainty set. This machine learning method is seamlessly integrated with adaptive robust optimization approach through a novel four-level optimization framework. This framework explicitly accounts for the correlation, asymmetry and multimode of uncertainty data, so it generates less conservative solutions than conventional robust optimization methods. Additionally, the proposed framework is robust not only to parameter variations, but also to data outliers. Because the resulting multi-level optimization problem cannot be solved directly by any off-the-shelf solvers, an efficient column-and-constraint generation algorithm is proposed to address the computational challenge. Applications on batch scheduling and on process network planning are presented to demonstrate the advantages of the proposed modeling framework and effectiveness of the solution algorithm.