We aim to answer following questions to combat the opioid epidemic: 1) Can we predict OD/OUD risks of patients in the future based on EHR history? We have developed machine learning (including deep learning) based predictive models using patients’ past EHR to predict future risk of OD and OUD. 2) Which regions or communities have most serious opioid problems for targeted interventions and optimized resource management? We have performed fine-grain geospatial analysis to discover disparities and geospatial patterns, and discover corrleted factors. 3) What are the opinions of the public, the emotions of the opioid users and the psychological effects of opioid use? We analyzed social media data through text mining (Twitter/Reddit) for in-depth studies.