IEPA 11 has ended
Back To Schedule
Wednesday, October 10 • 3:30pm - 3:40pm
Oral 18, Talk 5. "Detecting distress in adolescents and young adults using big data analysis of social media"

Sign up or log in to save this to your schedule, view media, leave feedback and see who's attending!

Feedback form is now closed.
Stefanie J Schmidt1,2, Danilo Croce3, Niels Bugge2, Chantal Michel2, Valentina Bellomaria3, Roberto Basili3, Frauke Schultze-Lutter4; 1Clinical Psychology for Children and Adolescents, University of Bern, Switzerland, 2University Hospital of Child and Adolescent Psychiatry and Psychotherapy, University of Bern, 3Enterprise Engineering, AI Research Group, University of Rome, Tor Vergata, Italy, 4Department of Psychiatry and Psychotherapy, Medical Faculty, Heinrich-Heine-University, Düsseldorf, Germany
The use of social media has grown exponentially in recent years, generating data that is a valuable source of information to detect potentially stressful events in an individual`s everyday-life. Thus, novel and interdisciplinary approaches are necessary to process this “big data” characterized by high volume, high velocity and high-variety information. Therefore, our study applied a combined support vector machine learning machine algorithm implemented within the Kernel-based Learning Platform (KeLP) and complex semantic language processing analysis to examine tweets written in English by adolescents and young adults between February 2017 and January 2018. Tweets were first classified according to 18 theory-derived life-event categories. Afterwards, a sentiment analysis was performed to identify a user`s attitudes towards this event (“positive”, “negative”, “neutral”, “ironic”). The dimension “experience” additionally captured each user`s emotional reaction related to this event (“distressful”, “helpful”, “neutral”). Furthermore, we examined potential gender-specific and socio-cultural differences. The automated classification process worked with sufficient accuracy of 76% and identified social relationships, hobbies/interests and interpersonal beliefs as the most prevalent events. Tweets related to mental health were experienced as being most distressful. Gender-differences were detected in that females tweeted more often about social and romantic relationships. With regard to socio-cultural differences, we identified primarily African-American Twitter-communities that more frequently discussed sociopolitical issues than other users. Thus, social big data mining is a promising analysis technique processing a huge amount of real-life data to identify stressors and supportive factors to promote mental health and well-being in adolescents and young adults.


Stefanie J Schmidt

University of Bern

Wednesday October 10, 2018 3:30pm - 3:40pm EDT
St. George CD Westin Copley Place, third floor

Attendees (7)