Àá½Ã¸¸ ±â´Ù·Á ÁÖ¼¼¿ä. ·ÎµùÁßÀÔ´Ï´Ù.
KMID : 1138720210470040149
Korean Public Health Research
2021 Volume.47 No. 4 p.149 ~ p.165
Construction and Evaluation of the Prediction Model for Smoking Cessation Behavior among Adult Smokers by Machine learning Algorithm
Choi Hwa-Young

Lee Kwang-Soo
Abstract
Objectives: The purpose of this study was to formulate a prediction model for smoking cessation behaviors of adult smokers using machine learning algorithms and to evaluate the effects of factors on the smoking behaviors.

Methods: Community health survey in 2019 was utilized, and the study subjects were persons over 19 years of age who smoke or have smoked. Dependent variables were smoking cessation behaviors such as smoking cessation attempts and successful smoking cessation. 84,126 persons had experiences of smoking cessation attempts, and 61,784 persons had experiences of successful smoking cessation. Prediction models were formulated by gender to prevent over-fitting. SAS 9.4 was utilized for the descriptive analysis, and SAS Enterprise-Miner 15.2 was utilized for the prediction model building and to visualize the scores.

Results: Study results showed that first, the prediction model using a random forest algorithm had an AUC(Area Under the Curve) of 0.7-0.8 in predicting smoking cessation attempts and successful smoking cessation, which indicates a stable result. Second, men who smoked for a longer period were less likely to attempt smoking cessation but were more likely to maintain their smoking cessation once they succeeded. Third, the likelihood for female smokers to quit smoking dropped quicker than their male counterparts as they smoked for a longer period. Female smokers who smoked longer were also less likely to succeed in smoking cessation.

Conclusion: This study suggested the possibility for machine learning algorithms to be adapted to smoking cessation program and suggested that sex and length of smoking periods will be major factors for supporting smoking cessation.
KEYWORD
Machine learning, Random forest, Smoking cessation behavior, Community health survey
FullTexts / Linksout information
Listed journal information
ÇмúÁøÈïÀç´Ü(KCI)