Analysis of Demographic, Familial, and Social Determinants of Smoking Behavior Using Machine Learning Methods

Abstract

Smoking behavior, encompassing both traditional tobacco and electronic cigarette use, is influenced by a range of demographic, familial, and social factors. This study examines the relationship between smoking habits and family dynamics through a cross-sectional survey of 100 participants, using an anonymous questionnaire to collect demographic data, smoking patterns, and familial interactions. Validated instruments, including the Penn State Electronic Cigarette Dependence Index and the Family Relationship Assessment Scale, were employed to assess smoking dependence and family dynamics. The analysis identified key patterns, such as increased smoking frequency among individuals experiencing higher family tension and variations in smoking habits across age and gender groups. Nocturnal smoking was linked to higher cigarette consumption, whereas early-day smokers exhibited a lower desire to quit. Machine learning models were applied to predict and classify smoking behaviors based on socio-demographic and familial variables, with an ensemble learning model achieving the highest accuracy (93.33%), outperforming k-nearest neighbors (90.00%), support vector machines (80.00%), and decision trees (83.33%). These findings underscore the complex interplay between family relationships and smoking behavior, providing insights for public health interventions. Additionally, this study highlights the potential of machine learning in behavioral research, demonstrating its utility in identifying and predicting smoking-related patterns.