Building upon an ongoing collaboration that has produced unique research resources describing the behavior of hundreds of young people at night time in public and private places in Switzerland (mobile phone sensors, log data, in-situ surveys, photos of drinks, videos of places, and semi-structured interviews), the proposed project will develop methods cutting across ubiquitous computing, addiction science, and human geography with three novel goals: (1) To understand the role of private spaces in youth nightlife and to identify discriminant features characterizing night time behavior of youth in private spaces including place ambiance, activities, and alcohol consumption, by integrating machine inference with quantitative and qualitative research. (2) To infer nightlife activities like nightclubbing vs. having a quiet night out based on sensor and video data and machine learning; and to dynamically characterize popular areas according to human nighttime behavior with additional social media and urban data sources. (3) To investigate the type and size of drinks in different contexts; the accuracy of event-level, self-reported alcohol consumption compared to manually coded information extracted from beverage photos; and the extent to which it is possible to infer attributes of the physical environment based on characteristics of the drink container and other discriminant cues from photos.