What is structure of water surrounding proteins remains as one of fundamental unsolved problems of science. Methods in biophysics only provide qualitative description of the structure and thus clarifying the collective phenomena of a huge number of water molecules is still beyond intuition in biophysics. We introduce a simulation-based data mining approach that quantitatively model the structure of water surrounding a protein as clusters of water molecules having similar moving behavior. The paper presents and explains how the advances of AI technique can potentially solve this challenging data-intensive problem.