Analysis of adolescent e-cigarette use and influencing factors based on machine learning algorithms

Objective: To understand the current situation of e-cigarette use and influencing factors among adolescents aged 15 and above in a certain city in Guangxi in order to provide data and references for controlling the prevalence of e-cigarettes among adolescents.Methods: A questionnaire survey was conducted among adolescents aged 15 and above in a certain city in Guangxi through multi-stage stratified cluster random sampling.Logistic regression, random forest, XGboost, support vector machine models, single hidden layer neural networks, and KNN models were applied comprehensively for the analysis of influencing factors.Results: The prevalence of e-cigarette use among adolescents aged 15 and above in a certain city in Guangxi was 1.68%, with the usage rates among high school and vocational high school students being 1.08% and 1.74%, respectively.Different machine learning models demonstrated varying levels of performance across evaluation metrics.Nine primary influencing factors were identified for adolescent e-cigarette use:exposure to e-cigarette advertisements on the internet in the past 30 days, friends'smoking habits, level of academic pressure, exposure to teachers smoking, depression status, gender, exposure to smoking in public places, perception of smoking as enhancing attractiveness among youth, and receiving free tobacco products.Conclusion: The prevalence of e-cigarette use among adolescents aged 15 and above in the city is relatively low.It is possible to combine the results of six machine learning models to predict adolescent electronic cigarette usage behavior and identify the characteristics of the user population.