A multi-stage predictive model for smoking cessation: Success and choices of medication approaches

Abstract

The number of deaths from diseases related to smoking is 11.6% of all Thai populations. So helping smokers to quit smoking cigarettes is one of the essential tasks of the medical personnel. This research developed a hybrid prediction model to support decision-making in the medical treatment of smoking cessation, which consists of medication decisions, the likelihood of three-month and six-month quitting, and medication choices using the data mining process. This research collected the treatment data from Thai Physicians Alliance Against Tobacco between 2015 to 2017 and was processed by data selection, data cleansing, data transformation, data resampling, and comparative experiments. Overall results were over 70% accuracy based on gradient boosted trees and neural network based on evolutionary parameter optimisation and ten-fold cross-validation evaluation method. Finally, the findings from the study would be beneficial to health personnel in making clinical decision support for better coverage of treatment for smokers.