A hybrid machine learning approach in predicting smoking behaviour: The case of Turkey

This study aims to analyze the smoking behaviour of people aged 15 and older in Turkey using supervised unsupervised machine learning methods. In this study, C4.5 Random Forest (RF) were trained predict behaviour, an apriori algorithm was used detect associations. Sensitivity, specificity, accuracy, positive predicted value (PPV), f-measure compare performances models. The Health Interview Survey 2019 with a sample size 17084 determine factors affecting smoking. Data analysis performance evaluation performed R programming language by RStudio. By association rules, gender, age, alcohol consumption are most representative attributes behaviour. Associations determined on smoking, non-smoking quit-smoking Also, it has been seen that RF better results than algorithm. It’s preferred use model, which had accuracy 0.909, specificity 0.965, sensitivity 0.782, PPV 0.908, 0.840 for predicting contributes literature covering comprehensive national health survey data methods Turkey. indicates can be such datasets.