Machine learning applications in tobacco research: a scoping review

Area of Science:

• Public Health
• Computer Science
• Data Science

Background:

• Tobacco research increasingly utilizes advanced analytical methods.
• Machine learning (ML) offers novel approaches to understanding and combating tobacco use.
• A comprehensive review of ML applications in tobacco control is needed.

Purpose of the Study:

• To identify and review the body of tobacco research literature that self-identified as using machine learning (ML) in the analysis.
• To categorize ML applications within tobacco research.
• To discuss implications and future directions for ML in tobacco control.

Main Methods:

• Searched MEDLINE, EMBASE, PubMed, CINAHL Plus, APA PsycINFO, and IEEE Xplore databases up to September 2020.
• Included peer-reviewed articles, dissertations, and conference papers using ML for empirical analysis of human tobacco experience; excluded genomics and imaging studies.
• Two reviewers screened studies, extracted data, and used narrative synthesis to classify findings into domains.

Main Results:

• Identified 74 studies employing ML in tobacco research.
• Studies were categorized into four domains: ML for smoking cessation (22), social media content analysis (32), smoker status classification (6), and outcome prediction (14).
• Trends in publication and application of ML methods were identified.

Conclusions:

• Machine learning (ML) is a powerful tool with significant potential to advance tobacco control research and policy.
• Further exploration of ML applications is warranted to enhance tobacco control efforts.
• The review highlights opportunities for ML researchers in the field of tobacco control.