Identifying current Juul users among emerging adults through Twitter feeds.

Area of Science:

• Computational social science
• Public health surveillance
• Machine learning applications

Background:

• Juul is a popular e-cigarette, raising concerns about uptake by non-smokers.
• Existing research often characterizes Juul-related social media content.
• Automated identification of Juul users is needed for targeted interventions.

Purpose of the Study:

• To assess the feasibility of identifying Juul users based on Twitter data.
• To apply machine learning models for automatic detection of Juul usage.

Main Methods:

• Collected survey data from 588 college students on Juul use and Twitter handles.
• Developed and tested supervised machine learning models (linear and deep learning).
• Utilized textual, social network (followers), and handcrafted features.

Main Results:

• A linear model incorporating textual and follower network features achieved 57% precision at 24% recall.
• Social network features significantly improved classification performance.
• The model can identify potential Juul users among college students on Twitter.

Conclusions:

• Social media data contains sufficient signals for supervised Juul use modeling.
• Machine learning models are valuable for focused public health intervention campaigns.
• This study supports automated surveillance of Juul use patterns and related behaviors via social media.