Classification of Twitter Users Who Tweet About E-Cigarettes

Area of Science:

• Social Media Analysis
• Public Health Surveillance
• Computational Social Science

Background:

• E-cigarette use has surged, with social media platforms like Twitter becoming key for information sharing and marketing.
• Understanding the diverse user types discussing e-cigarettes online is crucial for monitoring trends and health risks.
• Limited knowledge exists on the specific categories of users generating e-cigarette content on Twitter.

Purpose of the Study:

• To develop and demonstrate an automated methodology for classifying Twitter users discussing e-cigarettes.
• To categorize users into distinct types based on their online activity and profile information.

Main Methods:

• Collected over 11.5 million e-cigarette-related tweets from November 2014 to October 2016.
• Manually categorized a random sample of users into five types: individual, vaper enthusiast, informed agency, marketer, and spammer.
• Utilized machine learning algorithms, analyzing user metadata and tweeting behavior to build a classification model.

Main Results:

• Achieved an average F1 score of 83.3% in classifying five user types discussing e-cigarettes.
• High accuracy was observed for individuals (91.1%), informed agencies (84.4%), marketers (81.2%), and spammers (79.5%).
• Incorporating tweet-derived features significantly improved model performance by 10.6%, highlighting the importance of tweeting behavior analysis.

Conclusions:

• A novel method effectively classifies five distinct user types discussing e-cigarettes on Twitter.
• The model demonstrates high classification performance, particularly when incorporating tweeting behavior.
• Findings support public health surveillance, education, and regulatory efforts by identifying key online user groups.