Machine Learning-Based Nicotine Addiction Prediction Models for Youth E-Cigarette and Waterpipe (Hookah) Users

Area of Science:

• Public Health
• Data Science
• Adolescent Health

Background:

• Youth e-cigarette and hookah use is increasing despite health risks.
• Tailored cessation programs are crucial for adolescents.
• Understanding determinants of nicotine addiction in youth is vital.

Purpose of the Study:

• Identify social, mental, and environmental predictors of nicotine addiction in youth e-cigarette/hookah users.
• Develop machine learning models to predict nicotine addiction.
• Inform targeted interventions and policy.

Main Methods:

• Utilized data from 6511 participants in the National Youth Tobacco Survey (2019).
• Employed Random Forest with ReliefF and LASSO for prediction model development.
• Evaluated models using Root Mean Square Error (RMSE) and Confusion Matrix.

Main Results:

• Achieved high prediction performance for nicotine addiction in youth.
• Identified 193 significant predictor variables.
• Highlighted novel predictors like witnessing household e-cigarette use and perception of tobacco use.

Conclusions:

• Machine learning models effectively predict nicotine addiction in adolescent e-cigarette/hookah users.
• Identified predictors align with existing research and offer new insights.
• Findings can guide public awareness, youth education, and policy development for tobacco cessation.