Detection of adolescent periodic stress via micro-blog

摘要

Psychological stress detection via social media (micro-blog) is an emerging research topic, as it addresses one of the most common mental health issues, especially for teenagers who are not mature enough to deal with psychological pressures properly. Beyond simply detecting the stress category and level expressed in a single tweet, stress patterns, for example, periodic stress during a given time interval, often reveal high-level status of the user's suffering stress and hence make more sense. In this paper, we try to discover the periodicity of adolescent stress. Investigating fine-grained stressors, we first decompose the teenager's general stress series, and leverage a density-based clustering method to smooth discrete stress series points into the sequence of alternative stresson-stress intervals. Calculating the similarity between such intervals, the stress periodicity is finally identified by extending the symbol based DTW distance to sequences of stresson-stress intervals with the WARP algorithm. To the best of our knowledge, this is the first work detecting stress periodicity. Sufficient experiments upon real user study were conducted to evaluate both efficiency and effectiveness of our approach.